fix: handle llama.cpp context overflow in model adapter

[wefstratis-nv]

What does this PR do?

Handle llama.cpp context-overflow errors in the OpenAI-compatible model adapter.

Previously, llama.cpp HTTP 400 errors such as exceed_context_size_error were not recognized and caused evaluation failures. These errors now follow the existing context-length handling: return an empty completion with finish_reason="length".

The change covers both chat-completions and completions paths. Unrelated errors continue to propagate.

Validation:

• All 177 adapter tests pass, including 26 new regression cases.
• Pre-commit checks pass for the two changed files. The full --all-files command has not been run.
• Adapter-only coverage increased from 84.74% to 87.75%, but remains below the configured 95% threshold.
• No live benchmark validation has been performed.

Checklist

• I have read the contributing guidelines.
• The change is focused; unrelated "drive-by" edits are tracked as separate issues/PRs.
• Tests added or updated and pass locally, or N/A for docs-only / non-code changes (so CI unit/server checks pass when applicable).
• Pre-commit checks pass locally (pre-commit run --all-files) (so CI lint/format/copyright pass).
• All commits have DCO sign-off (git commit -s) (so the DCO check passes).

[copy-pr-bot]

This pull request requires additional validation before any workflows can run on NVIDIA's runners.

Pull request vetters can view their responsibilities here.

Contributors can view more details about this message here.

[ananthsub]

thanks @wefstratis-nv for the contribution! could you address the merge conflicts with the latest main so we can merge?