Picks up the doc-comment clarification from the copilot review on
DataDog/serverless-components#136 (https/fips features aren't
Cargo-enforced).

DataDog/serverless-components#136 merged at bb4dedeee20b949db3143c05e5a779b843a8a484.
The previous pin was the pre-merge branch SHA used during development.

Three real fixes from independent review:

1. **Cargo.lock dedup** — bump dogstatsd and datadog-fips from
   `aaac1a5d` to `bb4dedee` so they share the same serverless-components
   SHA as datadog-agent-config. Otherwise cargo resolves two distinct
   `dogstatsd` source entries (the dep transitively from agent-config
   pulled in a second copy at the newer rev). Source trees for both
   crates are byte-identical between the two SHAs; only #136 changed
   between them. PR description matches reality now.

2. **`capture_lambda_payload_max_depth` graceful deserializer** —
   the field was missing `#[serde(deserialize_with = ...)]`, violating
   the upstream `ConfigExtension::Source` contract. Without it, a
   malformed env value would silently reset *all* extension fields to
   their defaults (api_key_secret_arn, kms_api_key, …) and emit only
   a `tracing::warn!`. Adds `deser_opt_lossless` which returns `None`
   on bad input so the rest of the extension keeps its values.

3. **Test coverage gaps**:
   - YAML round-trip for the two source-to-config renames
     (`org_uuid → dd_org_uuid` and
     `lambda_customer_metrics_exclude_tags → custom_metrics_exclude_tags`).
     These were the exact cases where YAML behavior matters most and
     they only had env tests.
   - `FlushStrategy::EndPeriodically` (`"end,N"`) and
     `FlushStrategy::Continuously` (`"continuously,N"`) — the upstream
     enum has all four variants and bottlecap will rely on them
     behaviorally; previously only `End` and `Periodically` had
     coverage.

Test count: 33 → 37.

Picks up the doc-comment clarification from the copilot review on
DataDog/serverless-components#136 (https/fips features aren't
Cargo-enforced).

DataDog/serverless-components#136 merged at bb4dedeee20b949db3143c05e5a779b843a8a484.
The previous pin was the pre-merge branch SHA used during development.

Three real fixes from independent review:

1. **Cargo.lock dedup** — bump dogstatsd and datadog-fips from
   `aaac1a5d` to `bb4dedee` so they share the same serverless-components
   SHA as datadog-agent-config. Otherwise cargo resolves two distinct
   `dogstatsd` source entries (the dep transitively from agent-config
   pulled in a second copy at the newer rev). Source trees for both
   crates are byte-identical between the two SHAs; only #136 changed
   between them. PR description matches reality now.

2. **`capture_lambda_payload_max_depth` graceful deserializer** —
   the field was missing `#[serde(deserialize_with = ...)]`, violating
   the upstream `ConfigExtension::Source` contract. Without it, a
   malformed env value would silently reset *all* extension fields to
   their defaults (api_key_secret_arn, kms_api_key, …) and emit only
   a `tracing::warn!`. Adds `deser_opt_lossless` which returns `None`
   on bad input so the rest of the extension keeps its values.

3. **Test coverage gaps**:
   - YAML round-trip for the two source-to-config renames
     (`org_uuid → dd_org_uuid` and
     `lambda_customer_metrics_exclude_tags → custom_metrics_exclude_tags`).
     These were the exact cases where YAML behavior matters most and
     they only had env tests.
   - `FlushStrategy::EndPeriodically` (`"end,N"`) and
     `FlushStrategy::Continuously` (`"continuously,N"`) — the upstream
     enum has all four variants and bottlecap will rely on them
     behaviorally; previously only `End` and `Periodically` had
     coverage.

Test count: 33 → 37.