refactor: Fix implicit integer sign changes in strencodings #23756

This removes casts where they are nonsensical (ToUpper/ToLower) and adds them where needed. Then, remove the suppressions.

Also, reduce static memory usage.

• I checked that the first commit on my machine with gcc and clang produces the same binary with -O2.
• I checked that the second commit doesn't affect the binary size with clang, but reduces the size with gcc by 4096 Byte. I have not looked at the run-time memory change.

<!--e57a25ab6845829454e8d69fc972939a-->

The following sections might be updated with supplementary metadata relevant to reviewers and maintainers.

<!--174a7506f384e20aa4161008e828411d-->

Conflicts

Reviewers, this pull request conflicts with the following ones:

• #23595 (util: Add ParseHexstd::byte() helper by MarcoFalke)

If you consider this pull request important, please also help to review the conflicting pull requests. Ideally, start with the one that should be merged first.

crACK fa5865a9e339e7fe4068fbcce2ac055ed49eddfd

• The first commit prevents implicit sign change by manually casting the variables. This commit also removes unnecessary casting for ToUpper and ToLower functions.
• Since after the changes done in the first commit, we do not “expect” implicit-integer-sign-change warning, so it is removed from the sanitizer_suppresion/ubsan file.
• The second commit changes type of decode_table[256] from int to int8_t since all the elements of this array are in the range [-128, 127], so no need for a higher memory consuming data type (int)