Make globals const

[MaxKellermann]

This moves them from .data to .rodata and allows more compiler optimizations.

[MaxKellermann]

cl.exe rightfully complains that the const is duplicate; the code uses const in the typedef, which is an unorthodox thing to do... I'll fix it tomorrow.

[MaxKellermann]

Removed the redundant const. Let's see if cl.exe is now happy!

[Girgias]

This looks sensible to me, but would like the opinion of someone else.

[MaxKellermann]

cl.exe still isn't happy:

ext\exif\exif.c(507): error C2220: the following warning is treated as an error
ext\exif\exif.c(507): warning C4114: same type qualifier used more than once
ext\exif\exif.c(508): warning C4114: same type qualifier used more than once

Complaining about these two lines:

typedef const tag_info_type  tag_info_array[];
typedef const tag_info_type  *tag_table_type;

I don't quite get it - but I'll drop that hunk to get the rest merged.
In exchange, this force-push adds a new commit adding const to opcache's zend_shared_memory_handlers.

[MaxKellermann]

Spurious Travis failure again, the rest is green.

[alexdowad]

No objection on the mbstring part. I don't know what the maintainers of the other affected extensions think.

[alexdowad]

@MaxKellermann Thanks for all the cleanup you are doing.

[Girgias]

The MacOS build seems to be broken somehow in CI:

Run sudo make install
make: *** No rule to make target `/Users/runner/work/php-src/php-src/main/fopen_wrunner/work/php-src/php-src/main/fopen_wrappers.h', needed by `ext/standard/versioning.lo'.  Stop.

[MaxKellermann]

The MacOS build seems to be broken somehow in CI:

This looks like an unrelated spurious failure - this PR doesn't change anything with this header.

[Girgias]

The MacOS build seems to be broken somehow in CI:

This looks like an unrelated spurious failure - this PR doesn't change anything with this header.

ACK