Report-sink typed failure, opt-in report descriptor and run profile diagnostics, dependency-scratch storage fault code - #536

Preserve carried scratch exceptions and raise scratch_io_failed with no
arguments when a dependency scratch operation returns a bare I/O error, as
the exit-scratch consumer already does for the same status. Successful
operations and contract-error behaviour are unchanged.

Co-Authored-By: Claude Sonnet 5.5 <[email protected]>

…ailure

The default route installs an internal report sink. When the engine refuses
it (an unusable scratch arena descriptor, size or profile flag in the
environment, or a descriptor that is not a sealed memfd), the host reports a
generic internal inconsistency instead of the typed failure the engine
recorded. These rows run the real run_json.py against a real strategy module
with the arena environment set up per row, and expect report_sink_failed with
reason arena_backing:

- bad descriptor number, missing / empty / zero / negative / non-decimal /
  overflowing byte count, and an invalid jailed flag;
- descriptor 5 being a pipe, closed, a memfd with no seals, and a memfd with
  only some of the seals;
- a double whose setter fails with a recorded typed failure that is cleared
  when the strategy is freed: the failure must be read before the free.

Controls that must hold before and after: the default route with no arena
environment, a fully sealed memfd, and three doubles where the engine
recorded nothing (the generic internal-inconsistency line is kept).

Co-Authored-By: Claude Sonnet 5.5 <[email protected]>

…_io_failed

Co-Authored-By: Claude Sonnet 5.5 <[email protected]>

The entrypoint builds the extra run_json arguments from environment
variables. A caller that opens a descriptor for the report records itself has
no way to pass it through the container entrypoint. These rows drive the real
entrypoint with a private prefix and a logged run_json argv:

- variable unset: the argv is exactly the old one, for Pine and C++ input;
- variable set: the descriptor number is forwarded once as
  --report-sink-fd, an empty value is forwarded as an empty argument, and the
  position among the other forwarded options is fixed;
- invalid values (0, 1, 2, a negative number, 5, a non-number) are forwarded
  unchanged and refused by run_json itself with its own text; the entrypoint
  maps the refusal to its usual exit status 4.

A native row checks that the records of a run reach the caller's descriptor
and that standard output carries only the small document.

Co-Authored-By: Claude Sonnet 5.5 <[email protected]>

…rt sink is refused

When the engine refuses the internal report sink on the default route (for
example because the scratch arena descriptor, size or profile flag in the
environment is unusable), the host reported a generic internal
inconsistency and dropped the typed failure the engine had recorded. It now
reads that record, before the strategy state is freed, and reports it
(report_sink_failed with reason arena_backing for an unusable arena). When
the engine recorded nothing, the previous message is kept. No success path
changes.

Co-Authored-By: Claude Sonnet 5.5 <[email protected]>

Co-Authored-By: Claude Sonnet 5.5 <[email protected]>

…ailed arena was admitted

A success report carries only arena sizes, which show that an arena is
attached but not which profile ran or whether the jailed arena was admitted.
These rows expect two additive keys in the diagnostics object of every
executed success report (ordinary and selected windows, array output, path
and descriptor sinks): run_profile, the string "jailed" or "unjailed" fixed
when the run starts, and scratch_arena_jailed, a boolean that is true only
when the jailed profile ran and the native arena admission succeeded.

- the profile is the one frozen at startup, not a later environment read, and
  PF_SCRATCH_ARENA_JAILED alone never relabels an unjailed run;
- a real jailed run with an admitted sealed arena reads "jailed" / true, the
  same run without the profile reads "unjailed" / false;
- failure documents, capability documents and validate-only documents carry
  neither key;
- the replay converter keeps both keys and rejects wrong types;
- where a pinned test compares a whole document, it now asserts the two new
  keys first and then compares every other key exactly.

Co-Authored-By: Claude Sonnet 5.5 <[email protected]>

PINEFORGE_REPORT_SINK_FD=<n> makes the container entrypoint pass
--report-sink-fd <n> to run_json.py, for a caller that opened the descriptor
for the report records itself. The value is forwarded unchanged and run_json
remains the only validator, so its refusals (a number below 3, 5, a phase
descriptor, a non-number) apply as they do for the flag. With the variable
unset the entrypoint behaves exactly as before.

Co-Authored-By: Claude Sonnet 5.5 <[email protected]>

Co-Authored-By: Claude Sonnet 5.5 <[email protected]>

Co-Authored-By: Claude Sonnet 5.5 <[email protected]>

Co-Authored-By: Claude Sonnet 5.5 <[email protected]>

…e diagnostics

Every executed success report now carries two additive diagnostics keys:
run_profile ("jailed" or "unjailed", the profile fixed when the run started)
and scratch_arena_jailed (true only when the jailed profile ran and the
native scratch arena admission succeeded). The flag comes from the actual
admission result, not from the environment. The replay converter keeps both
keys. Failure, capability and validate-only documents are unchanged, as is
every existing key and its order. The documentation states what the marker
proves and what it does not (it is not an independent OS jail check).

Co-Authored-By: Claude Sonnet 5.5 <[email protected]>

Co-Authored-By: Claude Sonnet 5.5 <[email protected]>

…ries

Co-Authored-By: Claude Sonnet 5.5 <[email protected]>

…ries

Co-Authored-By: Claude Sonnet 5.5 <[email protected]>

Co-Authored-By: Claude Sonnet 5.5 <[email protected]>