[Bug #21931]

Keep tmp alive while RARRAY_CONST_PTR(tmp) is used by rb_str_format.

[alan: sunk the guard below usage]
Reviewed-by: Alan Wu <XrXr@users.noreply.github.com>

	[PATCH] search_nonascii(): Replace UB pointer cast with memcpy

	Casting a pointer to create an unaligned one is undefined behavior in C
	standards. Use memcpy to express the unaligned load instead to play by
	the rules.

	Practically, this yields the same binary output in many situations
	while fixing the crash in [Bug #21715].

[Bug #21671]

	[PATCH] [Bug #21629] Initialize `struct RString`

which appears to be missed in the previous commit for some reason.

	Suppress gcc 15 unterminated-string-initialization warnings

	Separate `__has_attribute` from `defined(__has_attribute)`

	Fix Visual C warnings:
	```
	regenc.h(121): warning C4067: unexpected tokens following preprocessor directive - expected a newline
	```

	Cast up `int` instruction code to `VALUE`

	Fix Visual C warnings:
	```
	iseq.c(3793): warning C4312: 'type cast': conversion from 'int' to 'void *' of greater size
	iseq.c(3794): warning C4312: 'type cast': conversion from 'int' to 'void *' of greater size
	```

	Do not let files depend on a phony target

	Detect `clock_gettime` and `clock_getres` for winpthreads

	[Bug #21380] Prohibit modification in String#split block

	Reported at https://hackerone.com/reports/3163876

[Bug #21172]

This fixes a rare CI failure.

The timeline of the race condition is:

- A `"foo" oid=1` string is interned.
- `"foo" oid=1` is no longer referenced and will be swept in the future.
- Another `"foo" oid=2` string is interned.
- `register_fstring` finds `"foo" oid=1`, but since it is about to be swept,
  removes it from `fstring_table` and insert `"foo" oid=2` instead.
- `"foo" oid=1` is swept, since it has the `RSTRING_FSTR` flag,
  a `st_delete` is issued in `fstring_table` which removes `"foo" oid=2`.

I don't know how to reproduce this bug consistently in a single test
case.

While profiling `strscan`, I noticed `rb_must_asciicompat` was quite
slow, as more than 5% of the benchmark was spent in it: https://share.firefox.dev/49bOcTn

By checking for the common 3 ASCII compatible encoding index first,
we can skip a lot of expensive operations in the happy path.