[V1][Minor] Minor simplification for get_computed_blocks

[WoosukKwon]

1. Early exit for the requests asking prompt_logprobs
2. If the last block is dropped, do not include it in prefix_cache_stats.queries.

[github-actions]

👋 Hi! Thank you for contributing to the vLLM project.

💬 Join our developer Slack at https://slack.vllm.ai to discuss your PR in #pr-reviews, coordinate on features in #feat- channels, or join special interest groups in #sig- channels.

Just a reminder: PRs would not trigger full CI run by default. Instead, it would only run fastcheck CI which starts running only a small and essential subset of CI tests to quickly catch errors. You can run other CI tests on top of those by going to your fastcheck build on Buildkite UI (linked in the PR checks section) and unblock them. If you do not have permission to unblock, ping simon-mo or khluu to add you in our Buildkite org.

Once the PR is approved and ready to go, your PR reviewer(s) can run CI to test the changes comprehensively before merging.

To run CI, PR reviewers can either: Add ready label to the PR or enable auto-merge.

🚀

[comaniac]

LGTM. It might cause some confusion: if users send a prompt with its length divisible by the block size multiple times, they will find the hit rate is not 100%. However, it shouldn't be a big deal.