fix: Throw TokenLimitReachedException on truncated tool calls

[chloe-pomegranate]

Summary

Fixes #193.

• When an API response hits the max token limit mid-way through generating a tool call (finish_reason=length with tool_calls present), throw TokenLimitReachedException instead of silently returning a FunctionCall with null/incomplete arguments.
• Add a secondary json_last_error() check in parseResponseChoiceMessageToolCallPart() to catch malformed tool call JSON regardless of finish reason.
• Exception includes the configured maxTokens value so callers can handle gracefully (e.g. increase the limit and retry).

Changes

src/Providers/OpenAiCompatibleImplementation/AbstractOpenAiCompatibleTextGenerationModel.php

• In parseResponseChoiceToCandidate(): detect finish_reason=length + tool_calls present in raw message data and throw TokenLimitReachedException before parsing the message.
• In parseResponseChoiceMessageToolCallPart(): replace silent json_decode with an explicit json_last_error() check that throws ResponseException on invalid JSON.

tests/unit/Providers/OpenAiCompatibleImplementation/AbstractOpenAiCompatibleTextGenerationModelTest.php

• testParseResponseChoiceToCandidateThrowsOnLengthWithToolCalls — truncated tool call throws exception.
• testParseResponseChoiceToCandidateTokenLimitIncludesMaxTokens — exception carries configured maxTokens (500).
• testParseResponseChoiceToCandidateLengthWithoutToolCallsDoesNotThrow — text-only + length still works.
• testParseResponseChoiceToCandidateToolCallsFinishReasonDoesNotThrow — normal tool calls unaffected.

Test plan

• vendor/bin/phpunit --filter "testParseResponseChoiceToCandidate" — all 12 tests pass.
• vendor/bin/phpunit — full suite passes (1059 tests, 3921 assertions).
• vendor/bin/phpstan analyse — no errors on source file.

Note: The Anthropic and Google providers would benefit from similar checks for their truncation signals (stop_reason=max_tokens and finishReason=MAX_TOKENS respectively).

[github-actions]

The following accounts have interacted with this PR and/or linked issues. I will continue to update these lists as activity occurs. You can also manually ask me to refresh this list by adding the props-bot label.

If you're merging code through a pull request on GitHub, copy and paste the following into the bottom of the merge commit message.

Co-authored-by: chloe-pomegranate <chloepomegranate@git.wordpress.org>
Co-authored-by: saarnilauri <laurisaarni@git.wordpress.org>

To understand the WordPress project's expectations around crediting contributors, please review the Contributor Attribution page in the Core Handbook.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 87.63%. Comparing base (30a1506) to head (b706802).

Additional details and impacted files

@@             Coverage Diff              @@
##              trunk     #214      +/-   ##
============================================
+ Coverage     87.45%   87.63%   +0.18%     
- Complexity     1194     1200       +6     
============================================
  Files            60       60              
  Lines          3858     3874      +16     
============================================
+ Hits           3374     3395      +21     
+ Misses          484      479       -5     

Flag	Coverage Δ
unit	87.63% <100.00%> (+0.18%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[chloe-pomegranate]

Pushed a new commit for coverage :)