ext/intl: Guard Spoofchecker restriction-level APIs at ICU 53 (fix build on ICU 50.x)

[GrahamCampbell]

PHP 8.4.22 fails to compile the intl extension when the system ICU is older than 53 — on Amazon Linux 2 (system ICU 50.2) the build breaks with USPOOF_ASCII, the other Spoofchecker restriction-level constants, and URestrictionLevel undeclared. The cause is GH-22055 (fixing #22053), which removed the #if U_ICU_VERSION_MAJOR_NUM >= 58 guards around the Spoofchecker restriction-level constants and the setRestrictionLevel() method, so they are now referenced unconditionally.

This is fine on master and PHP-8.5 (icu-uc >= 57.1) but not on PHP-8.4, which still accepts icu-uc >= 50.1. The symbols have existed since ICU 51 except USPOOF_SINGLE_SCRIPT_RESTRICTIVE (ICU 53, used in the setRestrictionLevel() body); since 50.2 satisfies the 50.1 minimum, configure passes and only the compile fails. ICU 53 is therefore the right floor: guarding at >= 53 builds on every ICU PHP-8.4 supports while still exposing the API on ICU 53+ — which is what GH-22055 wanted, since its >= 58 guard was too strict and hid the API on ICU 57.x.

The fix restores the guard at #if U_ICU_VERSION_MAJOR_NUM >= 53 around the constants and setRestrictionLevel() in spoofchecker.stub.php (regenerating spoofchecker_arginfo.h) and around the setRestrictionLevel() body in spoofchecker_main.c. The unrelated setAllowedChars() message change is kept.

[LamentXU123]

Indeed correct analysis. I wasn't aware PHP 8.4 has a different minimum ICU version than 8.5 and 8.6. Thank you for this PR!

[LamentXU123]

@GrahamCampbell We may need to add this into the PHP 8.4 NEWS file to notice others that we fix this.
This should only be merged into 8.4 so I will create 2 blank commit on upper branches (i.e. 8.5 and master)

[GrahamCampbell]

Added a news entry.

[LamentXU123]

@GrahamCampbell Thank you. The PR looks good and we will merge it if @devnexen has no opinions :)

[devnexen]

none I just wanted to see how you do :)

[LamentXU123]

Thanks!
(Crap. I forgot to add the author header in the commit... Sorry. But given the NEWS file the authorship is actually indicated. Lesson learnt :) )

[GrahamCampbell]

No worries. Thanks for the quick review.