migrate Accept, Content-Type, Link, MediaType, MediaRange to cats-parse

[novakov-alexey-zz]

migrate to cats-parse:

• Accept
• Content-Type
• Link
• MediaType
• MediaRange

Some tests fail due to few TODOs:

• Link header depends on Rfc3986Parser (will be done by someone else or in a separate PR)
• QValue parser from move ContentCoding, QValue and other headers to cats-parse #4013 needs to be merged first

Part of #3984

[novakov-alexey-zz]

There is one test case failing, I can't figure out why:

[error]   x Do a round trip through the Accept header (102 ms)
[error]    Accept: text/csv; quoteChar="'"; rowDelimiter=";"; escapeChar="\\\\\\\\"; charset="UTF-8"; columnDelimiter=" " != Accept: text/csv; quoteChar="'"; rowDelimiter=";"; escapeChar="\\\\"; charset="UTF-8"; columnDelimiter=" " (MediaTypeSpec.scala:55)
[error] org.http4s.MediaTypeSpec.$anonfun$new$16(MediaTypeSpec.scala:55)
[error] Actual:   ..."\\\\[\\\\]"; ch...
[error] Expected: ..."\\\\[]"; ch...

org.http4s.util.Writer#quote escapes every "\" with one more "\".
Failing test is parsing header from already parsed header, so that we get escapeChar="\\\\\\\\" instead of expected escapeChar="\\\\". @rossabaker do you have an idea why that may happen?

[rossabaker]

It looks like the parboiled2 version is returning a decoded quoted string, while the cats-parse version is returning the raw quoted string.

[novakov-alexey-zz]

It looks like the parboiled2 version is returning a decoded quoted string, while the cats-parse version is returning the raw quoted string.

We can try to unescape it only for MediaType parser.

[rossabaker]

I don't think Scala 3.0.0-M1 ever ran on JDK >= 14. We can ignore that error. While we're in flux, it would be nice to turn off the fail fast on CI. We're going to have a hard time getting any results here until we fix that.

[rossabaker]

I think I'd probably rather have the quotedString helper take care of unescaping, but I'd need to spend more time making sure they always get interpreted the same way.

Hard to say given the current state of CI on these port branches, but I think this unblocks us? I'm 👍, and it's not going to a prod branch yet, so there's time to clean up if we forgot something.

[novakov-alexey-zz]

Yes, I agree that unquoting could be done nicely via separate parser.