feat: add support for url_encode, url_decode, and try_url_decode

[parthchandra]

Which issue does this PR close?

Closes #4155

Rationale for this change

try_url_decode was returning incorrect results

What changes are included in this PR?

• Add native support for url_encode, url_decode, and try_url_decode (Spark 4.0+) by wiring up the existing datafusion-spark UDFs and handling the StaticInvoke serde for UrlCodec.encode / UrlCodec.decode.
  • Fix url_decode: try_url_decode (Spark 4.0) errors instead of returning NULL on malformed input #4155: try_url_decode was silently dropping the failOnError=false flag, causing Comet to error on malformed input where Spark returns NULL.

How are these changes tested?

SQL file tests -
expressions/url/url_decode.sql — valid inputs, malformed input errors
expressions/url/url_encode.sql — valid inputs, multibyte UTF-8
expressions/url/try_url_decode.sql — same inputs as url_decode, malformed input returns NULL

[andygrove]

Thanks @parthchandra, could you add the tests that I was adding in #4152 ? They covered some edge cases

[parthchandra]

Thanks @parthchandra, could you add the tests that I was adding in #4152 ? They covered some edge cases

Sure. Do we want to add parse_url to this PR as well?

[andygrove]

I don't mind either way. We can do that as separate PR if easier

[parthchandra]

Thanks @parthchandra, could you add the tests that I was adding in #4152 ? They covered some edge cases

Squashed and merged commits from #4152.

[parthchandra]

Note to reviewers: Spark's encode and url_encode are unrelated functions.
encode(string, charset) - converts a string to BINARY using a character set
`url-encode - percent encoding for URLs.

[comphead]

Suggested change

	("encode", UrlCodec.getClass) -> CometUrlEncodeStaticInvoke,
	("decode", UrlCodec.getClass) -> CometUrlDecodeStaticInvoke)
	("url_encode", UrlCodec.getClass) -> CometUrlEncodeStaticInvoke,
	("url_decode", UrlCodec.getClass) -> CometUrlDecodeStaticInvoke)

?

[andygrove]

Looking at the Spark source for UrlEncode / UrlDecode, the rewrite uses:

StaticInvoke(UrlCodec.getClass, dataType, "encode", Seq(child), ...)
StaticInvoke(UrlCodec.getClass, dataType, "decode", Seq(child, Literal(failOnError)), ...)

The third argument is the JVM method name on UrlCodec, which is literally "encode" and "decode". The user-facing SQL names url_encode / url_decode come from prettyName, not functionName.

[parthchandra]

@andygrove I merged your PR but parse_url from that PR failed in CI. I've backed it out, and it is tracked in #4156

[andygrove]

LGTM. user guide should also be updated to list these as supported. can be separate PR.

Thanks @parthchandra!