feat: route date_format through codegen dispatcher for non-native cases

[andygrove]

Which issue does this PR close?

Part of #3202.

Rationale for this change

Comet's date_format serde only emits the native to_char path for UTC sessions with a format literal in a small strftime-mappable whitelist. Every other case (non-UTC session, non-literal format, format outside the whitelist) falls back to Spark, which breaks up the Comet pipeline. Most production workloads run with a non-UTC session timezone, so the fallback is the common case in practice.

The Arrow-direct codegen dispatcher already in tree (CometScalaUDFCodegen, behind spark.comet.exec.scalaUDF.codegen.enabled) closure-serializes a bound Catalyst expression, ships it through a JvmScalarUdf proto, and Janino-compiles Spark's own doGenCode into a per-batch kernel that reads Arrow vectors and writes an Arrow output vector directly. DateFormatClass already has a real doGenCode (not CodegenFallback), is NullIntolerant, and has timeZoneId stamped on it by ResolveTimeZone during analysis. Routing it through the dispatcher therefore reproduces Spark behavior for any timezone or format, without a bespoke UDF class.

What changes are included in this PR?

Two commits.

fix(codegen) — Pre-existing bug surfaced while wiring date_format through the dispatcher. CometBatchKernelCodegen.defaultBody emitted this.col$ord.isNull(i) for every NullIntolerant input, but primitive Arrow vectors (TimeStampMicroTZVector, IntVector, Float8Vector, etc.) are wrapped in CometPlainVector at input-cast time and expose isNullAt, not isNull. Janino rejected the kernel with "method isNull not declared". emitTypedGetters already knew the right method name via nullCheckMethod. The fix exposes that helper so defaultBody picks the same name per column ordinal. New source test pins the chosen method for TimeStampMicroTZVector so the regression can't recur.

feat — Extract the closure-serialize + JvmScalarUdf emission from CometScalaUDF.convert into CometScalaUDF.emitJvmCodegenDispatch(expr, inputs, binding). CometDateFormat.convert keeps the native to_char path for UTC + whitelisted-format cases and now calls the helper for everything else. Gated by spark.comet.exec.scalaUDF.codegen.enabled (default false, marked experimental), so default behavior is unchanged. Non-UTC and non-whitelisted formats still fall back to Spark unless the user opts in.

The dispatcher's kernel cache is keyed on (serialized expression bytes, input column shapes), so distinct (format literal, timezone) combinations get distinct cached kernels just like a bespoke (format, tz) -> formatter cache would. No DateFormatUDF.scala is introduced.

How are these changes tested?

• CometTemporalExpressionSuite date_format tests: 10/10 pass. The three "falls back to Spark" tests are paired with a "routes via codegen dispatcher" sibling that enables the flag and asserts in-Comet execution. The disabled-flag variants still assert checkSparkAnswerAndFallbackReason with the dispatcher flag as the reason. date_format - timestamp_ntz input runs checkSparkAnswerAndOperator for every timezone under the codegen flag.
• CometSqlFileTestSuite date_format: 8/8 pass. date_format.sql now pins America/Los_Angeles and enables the codegen flag at file scope, with all queries plain query (no expect_fallback).
• CometCodegenSourceSuite: 30/30 pass, including the new NullIntolerant short-circuit uses isNullAt for CometPlainVector-wrapped columns regression test.
• CometCodegenSuite: 40+/40+ pass. No regressions in the dispatcher's existing surface.