Give a speed kick to combined is_serialized() + is_serialized_string() + maybe_unserialize() usage

[hakre]

In #17129 is_serialized_string() has been speed up by 23%. That was for the function itself.

However, that function is used in contexts that make the call to is_serialized() directly or indirectly via maybe_unserialize().

Most often these calls are redundant and must not be done at all.

If preconditions are already met, is_serialized_string() can be replaced with a simple check for the 's' or 'S' format and maybe_unserialize() can be replaced with @unserialize($data) as that's the code executed in maybe_unserialize() when is_serialze() had returned true on $data;

[hakre]

Added a patch. Technically it would allow to deprecate is_serialized_string() as it is not needed any longer by core.

Next to that I suggest to remove the trim() from is_serialized() as maybe_serialize() does not unserialize on trimmed but on untrimmed data. Technically it does fail on untrimmed data which has been reported as being serialized by is_serialize().

Related: #17375

[scribu]

@unserialize() is not an acceptable solution.

[c3mdigital]

See ​comment in #16504 regarding security implications.