fix: truncate oversized reranking passages instead of failing the request

[erichare]

What this PR does:

Changes the NVIDIA reranking request's truncate option from NONE to END.

With NONE, a single query+passage pair over the model's 8192-token limit fails the whole batch with HTTP 400 (Input length 16384 exceeds maximum allowed token size 8192), and because rerank batches are dispatched fail-fast, the entire findAndRerank command fails — deterministically, for any query whose candidate set contains one oversized document. No hybridLimits setting avoids it when the oversized document ranks highly.

Observed against a production collection containing large OCR'd table chunks (13–15k chars, digit/punctuation-heavy): ~3% of eval queries fail 100% of the time with RERANKING_PROVIDER_CLIENT_ERROR. With END, the NIM truncates the oversized pair and scores the remaining tokens — a strictly better outcome than failing the whole command.

One-line constant change plus javadoc; no config surface added while findAndRerank is still stabilizing. If per-model configurability is wanted later, truncate can move into reranking-providers-config.yaml as a follow-up.

Validation:

• NvidiaRerankingProviderTest + RerankingProviderTest pass (4 tests)
• formatting clean (fmt:format)

Which issue(s) this PR fixes:

N/A — follow-up to the GPU-plane concurrency investigation (same series as #2530).

Checklist

• Changes manually tested after deployment
• Automated Tests added/updated (no existing test exercises the wire payload; not adding mock-server scaffolding for a constant)
• Documentation added/updated (javadoc)
• CLA Signed: DataStax CLA

[github-actions]

Unit Test Coverage Report

Overall Project	52.97%	🍏

File	Coverage
NvidiaRerankingProvider.java	17.58%	🍏

[github-actions]

Integration Test Coverage Report (dse69-it)

Overall Project	71.43%	🍏

File	Coverage
NvidiaRerankingProvider.java	12.73%	🍏

[github-actions]

Integration Test Coverage Report (hcd-it)

Overall Project	72.75%	🍏

File	Coverage
NvidiaRerankingProvider.java	12.73%	🍏