Chg #960: Pass non-null and non-expression values only to dbTypecast()

[KalimeroMK]

Fix #960

Changes

• null values are now passed through without calling ColumnInterface::dbTypecast(). The check is inlined at the call sites in AbstractDMLQueryBuilder (insert/update/batch) and AbstractColumnDefinitionBuilder::buildDefault().

• No new API. The first version of this PR added a dbTypecast() wrapper function, but a wrapper that only filters null adds call overhead for no measurable gain (see benchmark), so the check was inlined instead.

• DateTimeValueBuilder is untouched: DateTimeValue::$value is non-nullable by type.

• Column implementations are intentionally left untouched. Two findings from the driver test matrices:

  • Expression is Stringable, so removing the ExpressionInterface arms silently stringifies expressions instead of passing them through (breaks DDL builders in db-pgsql/db-mssql that pass Expression default values directly).
  • Some driver implementations process expressions: Yiisoft\Db\Mssql\Column\BinaryColumn::dbTypecast() unwraps Param values into CONVERT(VARBINARY(MAX), ...) expressions, so expressions must keep reaching dbTypecast().

  Net effect: only null is safe to pre-filter. Simplifying the implementations needs a coordinated change across the driver packages and can be done as a follow-up for 3.0.

Benchmark

2M calls per round, median of 7 rounds, PHP 8.5 with OPcache + JIT (tracing), workload: 80% scalars, 10% null, 10% expressions, IntegerColumn + StringColumn:

approach	time per round
current master (direct call)	~34 ms
wrapper function (first version of this PR)	~42 ms
check inlined at the call site (this PR)	~34.5 ms

The inlined check is within noise of master; the wrapper function added ~4 ns per call on trivial casts.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 98.62%. Comparing base (f0b0d2b) to head (82017d6).
⚠️ Report is 1 commits behind head on master.

Additional details and impacted files

@@            Coverage Diff            @@
##             master    #1197   +/-   ##
=========================================
  Coverage     98.62%   98.62%           
- Complexity     1645     1649    +4     
=========================================
  Files           120      120           
  Lines          4288     4289    +1     
=========================================
+ Hits           4229     4230    +1     
  Misses           59       59           

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[Tigrov]

But it seems the function that only checks for null doesn't make sense.

#960 (comment)

[KalimeroMK]

Agreed that a null-only wrapper doesn't pay for itself — dropped the function and inlined the null check at the call sites in 82017d6. Benchmark is at parity with master now.