[IR] Don't verify module flags on every access (llvm#102153)

8b4306c introduced validity checks for every module flag access,
because the auto-upgrader uses named metadata before verifying the
module.

This causes overhead for all other accesses, and the check is, in fact,
only need at that single place. Change the upgrader to be careful when
accessing module flags before the module is verified and remove the
checks on all other occasions.

There are two tangential optimizations included: first, when querying a
specific flag, don't enumerate all other flags into a vector as well.
Second, don't use a Twine for getNamedMetadata(), which has
materialization overhead -- all call sites use simple strings that can
be implicitly converted to a StringRef.

Change-Id: I724c67e52f080caba6bc16ec983f6118429408a1

Author: aengelke

revert_patches.txt

@@ -103,10 +103,6 @@ dfeb3991fb48 Remove the `x86_mmx` IR type. (#98505)
 b7e4fba6e5dc Cleanup x86_mmx after removing IR type  (#100646) (Reason: dependent on dfeb3991fb48)
 Ron: still broken 9-6-24
 ---
-Revert: causing issues with IRVerifiery heterodebug
-[IR] Don't verify module flags on every access (#102153)
-Ron
----
 revert:  1ca9fe6db334  Reapply "[Attributor][AMDGPU] Enable AAIndirectCallInfo for AMDAttributor
 ---
 revert : breaks build of amdgpu flat intrinsics