File: Re-add support to skip CR (\r) in File::get_as_text

[akien-mga]

This was removed in #63481, and we confirmed that it's better like this,
but we add back the possibility to strip CR as an option, to optionally
restore the previous behavior.

For performance this is done directly in String::parse_utf8.

Also fixes Android FileAccess::get_line() as this one should strip CR.

Supersedes #63717.

Will backport for 3.x with the get_as_text() option defaulting to true to fix the compat breakage.

@bruvzg My changes to String::parse_utf8() might be a huge hack, please review.

---

Speed comparison on a big file with a debug build on Linux, as described in #63717 (comment):

• Original File::get_as_text() from before Address remaining scoped storage regressions #63481

Unix file: 3.107707 s
DOS file:  3.214315 s

• Current master after Address remaining scoped storage regressions #63481

Unix file: 0.136303 s
DOS file:  0.162586 s

• PR File: Remove \r in get_as_text() to keep standardized Unix format #63717 with replace("\r", "")

Unix file: 0.180824 s
DOS file:  1.624483 s

• This PR

Unix file: 0.129991 s
DOS file, keep CR: 0.148745 s
DOS file, skip CR: 0.190492 s

[bruvzg]

parse_utf8 changes seems fine.

[akien-mga]

I still had a doubt about my parse_utf8 changes so I added some tests, and indeed they fail :P

[bruvzg]

Test files seem to be wrong, CRLF one it only has LF.

[akien-mga]

Test files seem to be wrong, CRLF one it only has LF.

Indeed, my .gitattributes override doesn't seem to work as intended. The Git documentation is horrible to read :|

Edit: I finally got it to work.

---

With your advice I could fix the FileAccess test but the String one still fails.

./tests/core/string/test_string.h:155:
TEST CASE:  [String] UTF8 with CR

./tests/core/string/test_string.h:166: ERROR: CHECK( no_cr == base.replace("\r", "") ) is NOT correct!
  values: CHECK( H == Hello darkness
My old friend
I've come to talkWith you again )

[DarkKilauea]

Should an enumeration that selects the desired line ending be used instead? I understand this is a quick fix, but I'm concerned about adding a parameter that will need to be changed later on, resulting in a breaking change down the line.

enum LineEnding {
    LineEnding_Unix, // lf
    LineEnding_Mac // cr
    LineEnding_Windows // crlf
}

[akien-mga]

Probably overkill, no modern platform uses CR only anymore. It was used on Mac OS prior to Mac OS X.
Since Mac OS X (2001), it's using LF like other Unixes. Windows uses CRLF.