…e is_account_client docstrings (#1317)

## 🥞 Stacked PR
Use this
[link](https://github.com/databricks/databricks-sdk-py/pull/1317/files)
to review incremental changes.
-
[**stack/use-test-environment-type-in-integration-tests**](#1317)
[[Files
changed](https://github.com/databricks/databricks-sdk-py/pull/1317/files)]
-
[stack/resolve-host-metadata-on-init](#1318)
[[Files
changed](https://github.com/databricks/databricks-sdk-py/pull/1318/files/0eff155d0ca7fc796d878cf9ee332dac5149be6a..63156f63eb80992d0c25a7b341d0b19b77dc9527)]
-
[stack/cloud-from-host-metadata](#1320)
[[Files
changed](https://github.com/databricks/databricks-sdk-py/pull/1320/files/63156f63eb80992d0c25a7b341d0b19b77dc9527..f9b0c3e2df72ddecb0fccb0d59af46f3a06846d2)]
-
[stack/resolve-token-audience-from-host-metadata](#1321)
[[Files
changed](https://github.com/databricks/databricks-sdk-py/pull/1321/files/f9b0c3e2df72ddecb0fccb0d59af46f3a06846d2..d97bd75a5692faddc36d51d46208513e442e7c87)]
-
[stack/google-auth-use-host-metadata](#1322)
[[Files
changed](https://github.com/databricks/databricks-sdk-py/pull/1322/files/d97bd75a5692faddc36d51d46208513e442e7c87..87b4c1a35e3fe1abfb125069ba1590eeaecf34e9)]

---------
## Summary

Replace `is_account_client` heuristics and `DATABRICKS_ACCOUNT_ID`
presence checks in integration test fixtures with explicit
`TEST_ENVIRONMENT_TYPE`-based filtering. Also deprecate
`is_account_client` in docstrings across `Config` and `ApiClient`.

## Why

Integration test fixtures in `tests/integration/conftest.py` used two
different heuristics to decide whether a test should run:

1. `account_client.config.is_account_client` — inspects the host URL to
infer whether the SDK is talking to an account-level API. This is
unreliable now that unified hosts can serve both account and workspace
APIs, and the property already raises `ValueError` when
`experimental_is_unified_host` is set.
2. `"DATABRICKS_ACCOUNT_ID" in os.environ` — a proxy check used to skip
workspace fixtures on account environments. This is fragile and doesn't
express intent clearly.

The test orchestrator now sets `TEST_ENVIRONMENT_TYPE` explicitly to one
of `ACCOUNT`, `UC_ACCOUNT`, `WORKSPACE`, or `UC_WORKSPACE`. Using this
value directly makes fixture skip-guards authoritative and easy to
reason about.

## What changed

### Interface changes

None. No public SDK API changes.

### Behavioral changes

None for SDK users. Integration test fixture skip conditions are now
driven by `TEST_ENVIRONMENT_TYPE` instead of `is_account_client` and
`DATABRICKS_ACCOUNT_ID`.

### Internal changes

- **`tests/integration/conftest.py`**: All four client fixtures (`a`,
`ucacct`, `w`, `ucws`) now check `TEST_ENVIRONMENT_TYPE` against an
explicit allowlist instead of using `is_account_client` or
`DATABRICKS_ACCOUNT_ID`.
- **`databricks/sdk/config.py`**: Updated `is_account_client` docstring
to note deprecation — clients can now support both workspace and account
APIs from a single host.
- **`databricks/sdk/core.py`**: Added matching deprecation note to
`ApiClient.is_account_client`.

## How is this tested?

The changed code is test infrastructure itself. Correctness will be
validated when integration tests run against environments that set
`TEST_ENVIRONMENT_TYPE`.

NO_CHANGELOG=true

#1318)

## 🥞 Stacked PR
Use this
[link](https://github.com/databricks/databricks-sdk-py/pull/1318/files)
to review incremental changes.
-
[**stack/resolve-host-metadata-on-init**](#1318)
[[Files
changed](https://github.com/databricks/databricks-sdk-py/pull/1318/files)]
-
[stack/cloud-from-host-metadata](#1320)
[[Files
changed](https://github.com/databricks/databricks-sdk-py/pull/1320/files/cd35488c886dc7a57ff3b1f79d70d1089a3ab6cf..e1872e347ea0b17197154389a80381d84d99a210)]
-
[stack/resolve-token-audience-from-host-metadata](#1321)
[[Files
changed](https://github.com/databricks/databricks-sdk-py/pull/1321/files/e1872e347ea0b17197154389a80381d84d99a210..51c62a915c2a3d0bd3c5ad720c26c5c79647976e)]
-
[stack/google-auth-use-host-metadata](#1322)
[[Files
changed](https://github.com/databricks/databricks-sdk-py/pull/1322/files/51c62a915c2a3d0bd3c5ad720c26c5c79647976e..9898fc156ef3d0cadc3c254e033cfaf20caa537d)]

---------
## Summary

Call `_resolve_host_metadata()` automatically during `Config.__init__`
so that `account_id`, `workspace_id`, `discovery_url`, and `cloud` are
populated from the `/.well-known/databricks-config` endpoint before auth
is initialized. Metadata fetch failures are non-fatal — a warning is
logged and initialization continues with whatever the user has
explicitly configured.

## Why

`_resolve_host_metadata()` already existed but had to be called
explicitly by callers who wanted auto-discovery. This meant that
`WorkspaceClient` and `AccountClient` did not benefit from it out of the
box. Beyond the missing call, the old implementation was also strictly
fatal: any HTTP error or missing field (`account_id`, `oidc_endpoint`)
raised a `ValueError`, making it unsafe to run automatically since not
all hosts expose the endpoint.

This PR wires up the call during `Config.__init__` (between
`_fix_host_if_needed` and `_validate`) and hardens the implementation so
it degrades gracefully: HTTP failures produce a warning, and missing
optional fields are simply left unset. The only remaining hard failure
is when a `{account_id}` template in `oidc_endpoint` cannot be
substituted because no `account_id` is available — this is a
configuration error that should still surface.

Auto-discovery is currently gated to hosts with
`experimental_is_unified_host=True` (a `# TODO: Enable this everywhere`
marks the guard) to avoid unintended network calls on standard workspace
and account deployments until the endpoint is broadly available.

## What changed

### Interface changes

None. `_resolve_host_metadata()` remains private.

### Behavioral changes

For hosts with `experimental_is_unified_host=True`: `Config.__init__`
now makes an HTTP request to `{host}/.well-known/databricks-config` and
silently back-fills `account_id`, `workspace_id`, `discovery_url`, and
`cloud` from the response. Previously this did not happen.

All other hosts: no change.

### Internal changes

- **`databricks/sdk/config.py`**: Added `_resolve_host_metadata()` call
to `Config.__init__`. Added `HostType.UNIFIED` guard. Made HTTP failures
non-fatal (warning + return). Removed mandatory-field assertions
(`account_id is required`, `discovery_url is required`) — missing fields
are now silently skipped.
- **`tests/conftest.py`**: Added autouse `stub_host_metadata` fixture
that mocks `get_host_metadata` for all unit tests, preventing unexpected
network calls.
- **`tests/test_config.py`**: Replaced the old `requests_mock`-based
tests (which called `_resolve_host_metadata()` directly) with
`mocker.patch`-based tests that exercise the full `Config.__init__`
path. Added `test_resolve_host_metadata_skipped_for_non_unified` to
assert the guard works.

## How is this tested?

Unit tests in `tests/test_config.py` cover: successful resolution of all
fields, no-overwrite of pre-configured values, HTTP failure tolerance,
missing `oidc_endpoint`, and the guard that skips non-unified hosts.

NO_CHANGELOG=true

---------

Co-authored-by: Claude Sonnet 4.6 <[email protected]>