[Phoenix500526]

Important

Fork CI setup: For pull requests from a fork to main, enable workflows in your fork's Actions tab before you push.
See the fork CI setup guide.

Which issue does this PR close?

• Closes APPROX_MEDIAN with ORDER BY incorrectly treats the sort key as a percentile argument #27157.

Rationale for this change

APPROX_MEDIAN(x ORDER BY k) incorrectly passes the sort key to the shared percentile accumulator, which treats it as a percentile argument. For an integer sort key, this causes:

What changes are included in this PR?

• Declare APPROX_MEDIAN as order-insensitive so the framework discards unnecessary aggregate ORDER BY keys.
• Add an SLT regression test for the reported query.

What is the testing strategy for this PR?

• Confirmed the regression test fails with the cached pre-fix test executable and passes with the fix.
• Passed aggregate.slt and the workspace extended test suite.
• Passed cargo fmt --all and cargo clippy --all-targets --all-features -- -D warnings.

Are there any user-facing changes?

Queries such as the following now return 2.0 instead of an error:

SELECT approx_median(x ORDER BY k)
FROM (VALUES (1.0, 3), (2.0, 1), (3.0, 2)) t(x, k);