pcre2: Update to upstream version 10.42 (take two)

[akien-mga]

Changelog: https://github.com/PCRE2Project/pcre2/blob/pcre2-10.42/ChangeLog

This fixes support for RISC-V architectures in the sljit library, so we enable the module's compilation for rv64 too.

Redo of #70447 which I reverted due to a regression described there:
#70447 (comment)

Fixes #76174

---

From @vonagam:

This PR caused those lines to appear (repeated multiple times) in unit tests logs at the start.

ERROR: Condition "p_ptr == nullptr" is true.
   at: free_static (core/os/memory.cpp:148)

While tests do not fail because of that, this does not seem right...
Or is it unavoidable known issue?

My early findings:

Backtrace:

(gdb) bt
#0  Memory::free_static (p_ptr=0x0, p_pad_align=false) at core/os/memory.cpp:149
#1  0x0000000006c247fb in _regex_free (ptr=0x0, user=0x0) at modules/regex/regex.cpp:43
#2  0x0000000006c1cb84 in pcre2_match_32 (code=0xc9019b0, subject=0xc8a74c0, length=9, start_offset=0, options=0, match_data=0xc830060, mcontext=0xc9863b0) at thirdparty/pcre2/src/pcre2_match.c:6846
#3  0x0000000006c25c85 in RegEx::search (this=0x7fffffffc300, p_subject=..., p_offset=0, p_end=-1) at modules/regex/regex.cpp:219
#4  0x00000000057cfee7 in TestRegEx::DOCTEST_ANON_FUNC_6556 () at ./modules/regex/tests/test_regex.h:81
#5  0x00000000053d282d in doctest::Context::run (this=0x7fffffffd0e0) at ./thirdparty/doctest/doctest.h:6930
#6  0x00000000057d5540 in test_main (argc=2, argv=0x7fffffffd728) at tests/test_main.cpp:174
#7  0x0000000005393ec2 in Main::test_entrypoint (argc=2, argv=0x7fffffffd728, tests_need_run=@0x7fffffffd1bf: true) at main/main.cpp:608
#8  0x0000000005333cc2 in main (argc=2, argv=0x7fffffffd728) at platform/linuxbsd/godot_linuxbsd.cpp:55

It probably comes from this change:

9. Removed the use of an initial backtracking frames vector on the system stack
in pcre2_match() so that it now always uses the heap. (In a multi-thread
environment with very small stacks there had been an issue.) This also is
tidier for JIT matching, which didn't need that vector. The heap vector is now
remembered in the match data block and re-used if that block itself is re-used.
It is freed with the match data block.

We likely need to change our _regex_free (and maybe _regex_malloc) to take this new heap vector into account somehow.

[akien-mga]

Added a nullptr check in _regex_free as suggested by @lawnjelly, and it solved the error spam.

I don't know if there are other implications for that upstream change which caused this error spam in the first place.

[akien-mga]

For 4.0 and 3.5, I'm making a simpler PR that just adds the null check, to fix #76174: #78137.

[RedworkDE]

While the pcre2 docs don't appear to specify anything about the required behavior of the passed free function, the standard C free function is explicitly specified to do nothing when passed a null pointer, so I would assume this is just caused by some simplifications and them removing some otherwise unnecessary null checks.

[akien-mga]

Cherry-picked for 3.6.

[aaronfranke]

As of this PR cherry-picked into the 3.x branch, I am unable to compile Godot on macOS, because Godot 3.x targets macOS 10.15+, and this change breaks support for macOS < 11.0.

If I revert this PR, it compiles fine (fails on linking but that's a different problem).

[bruvzg]

I am unable to compile Godot on macOS, because Godot 3.x targets macOS 10.15+, and this change breaks support for macOS < 11.0.

Target version should be changed to 11.0, 10.15 for arm64 existed only as the early alphas.

[aaronfranke]

Thanks, fixed in #82007