fix: improve delete_metrics_splits error handling and doc comments

[g-talbot]

Summary

• Rewrite delete_metrics_splits in postgres metastore to distinguish "not found" (warn + succeed) from "not deletable" (FailedPrecondition), matching the non-metrics delete_splits pattern
• Fix row_keys_proto doc comment to use plain code span instead of broken rustdoc link

Split out from #6247 (docs PR) where these code changes didn't belong.

Test plan

• Existing metastore tests pass
• Postgres delete tests cover not-found and not-deletable cases

🤖 Generated with Claude Code

[mattmkim]

@codex review

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: fcf8e650a7

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

Lock metrics rows before deriving not_deletable state

delete_metrics_splits computes input_splits from metrics_splits without FOR UPDATE, unlike the non-metrics delete_splits flow it is mirroring. Under concurrent mark_metrics_splits_for_deletion updates, this statement can evaluate an older split_state snapshot and return FailedPrecondition for splits that were already transitioned to MarkedForDeletion by a racing transaction, causing spurious delete failures and retry churn. Adding row locking in the state-read CTE (as done in delete_splits) avoids this stale-state race.

Useful? React with 👍 / 👎.