fix(optimizer): coerce numeric types in index column hints for int-float comparisons

[Tyagiquamar]

What problem does this PR solve?

Fixes #6172.

When an INT property is compared to an integer-valued floating-point literal (e.g. 3000000000.0), queries executed via index scans on that property previously dropped matching rows. This occurred because OptimizerUtils::appendColHint used the float Value directly in IndexColumnHint without coercing it to the column definition schema type (INT64). In RocksDB storage, index keys for INT and DOUBLE use distinct encodings, causing index prefix seek misses.

What is changed and how does it work?

1. Introduced SchemaUtil::coerceValue to safely convert representable integer floating-point literals to integer values for integer property types (and integer literals to float for float property types).
2. Exposed OptimizerUtils::normalizeFilterValue to normalize filter expressions before creating column hints.
3. Updated OptimizerUtils::appendColHint to validate and coerce candidate filter values to the column's expected data type for both prefix and range scans.
4. Added test cases in IndexScanRuleTest for exact-integer float literals on int columns and int literals on double columns.
5. Added test coverage in LookupValidatorTest.

Check List

Tests:

• Unit test added in IndexScanRuleTest
• Unit test added in LookupValidatorTest

[CLAassistant]

All committers have signed the CLA.