ANSI mode test coverage: re-enable stale ignored tests and add missing cases #5076

What is the problem the feature request solves?

An ANSI-mode audit against Spark 4.1.1 found several test-coverage gaps around ANSI behavior. Closing them locks in the behavior that is already correct and captures the known divergences as regression tests.

Stale disabled tests to re-enable:

• spark/src/test/resources/sql-tests/expressions/array/element_at_ansi.sql and array/get_array_item_ansi.sql are entirely query ignore(#3375) with comments claiming "Comet throws: different error message". ANSI mode array access error messages don't match Spark format #3375 is closed: the error-message work merged (native GetArrayItem re-enabled shortly after), and CometArrayExpressionSuite now asserts exact Spark message parity for both expressions. These files should be un-ignored.

Missing ANSI SQL file tests (compare the existing math/abs_ansi.sql, math/arithmetic_ansi.sql, url/parse_url_ansi.sql pattern):

• round: no ANSI case at all; also no integral negative-scale overflow case in any mode.
• conv: no ANSI overflow case (e.g. conv('FFFFFFFFFFFFFFFFF', 16, 10) throws under ANSI).
• elt: no out-of-range index case, no ANSI config.

Specific untested inputs surfaced by the audit (each is the repro for a filed bug; add as query ignore(<issue-url>) until fixed, then un-ignore):

• Long.MinValue div -1 under ANSI (IntegralDivide overflow).
• CAST(1.0 AS DOUBLE) % CAST(0.0 AS DOUBLE) under ANSI (float remainder by zero).
• CAST(true AS DECIMAL(1,1)) in legacy/try mode (boolean-to-decimal overflow).
• round(5000000000000000000L, -19) in both modes.

Assertion strength: several suites assert only on message substrings, which is why wrong-exception-type divergences pass today (decimal div-by-zero, next_day, make_date, decimal-to-decimal cast). Where the error class is expected to match, assert on the exception type / getCondition() rather than substrings, and remove the decimal-to-decimal special case at CometCastSuite.scala:2402-2405 once the underlying issue is fixed.

Additional context

Found by an ANSI-mode audit of all native expressions against Spark 4.1.1.