fix(mito2): schema-safe skipping index pruning

[fengys1996]

I hereby agree to the terms of the GreptimeDB CLA.

Refer to a related PR or issue link (optional)

#8074 but skipping index.

What's changed and what's your intention?

This pr mainly change

• Made skipping-index pruning schema-safe by checking predicate type compatibility per SST, preventing false negatives under schema evolution.
• Switched from whole-applier fallback to column-subset fallback, so compatible predicates still prune while incompatible ones are skipped.

PR Checklist

Please convert it to a draft if some of the following conditions are not met.

• I have written the necessary rustdoc comments.
• I have added the necessary unit tests and integration tests.
• This PR requires documentation updates.
• API changes are backward compatible.
• Schema or data changes are backward compatible.

[gemini-code-assist]

Code Review

This pull request introduces logic to handle column type changes in bloom filter indexing by verifying type compatibility between region metadata and SST metadata. It adds an SstApplyPlan to manage predicates and ensures that indexes are only applied when types match, preventing issues after an ALTER TABLE operation. Review feedback suggests optimizing the SstApplyPlan by excluding columns that are entirely missing from the SST to improve cache key consistency and reduce unnecessary map entries.

[evenyag]

The comment is outdated.

[evenyag]

We use a flag has_type_mismatch and reuse the plan in the inverted applier. Should we follow the same pattern?

greptimedb/src/mito2/src/sst/index/inverted_index/applier.rs

Lines 387 to 389 in f1ad472

	if !has_type_mismatch {
	return Ok(Some(self.default_plan.clone()));
	}