std.ascii: remove LUT and deprecations

[wooster0]

I wrote this after #12448 got merged but decided to wait for 0.10.0 to be released before pulling it. Now I've decided to pull it a bit earlier since 0.10.0 is just about to be released, anyway (and in case someone else would accidentally start working on the same thing or something).

This should be merged after 0.10.0 and removes all deprecations, updates the codebase, does some other cleanups etc., and also removes the LUT (Look-Up Table): this was already discussed quite a bit in #11629 (see #11629 (comment) etc.) and I've also discussed this with @topolarity in more detail after that, but basically the reason is that the LUT made us depend on cache performance. It also reduces binary size. See also the commit's description for more detail.

Removing both the deprecations and the LUT at the same time is a very convenient timing, by the way. It shouldn't make review that much harder, I hope.

[IntegratedQuantum]

I think I found something:

/// Returns whether the character is printable and has some graphical representation.
/// This also returns `true` for the space character.
/// This is the same as `!isControl(c)`.
pub fn isPrint(c: u8) bool {

This is the same as !isControl(c).

This was not true in the lookup-table implementation.
The lookup-tables always had a zero for non-ascii characters.
And I guess some part of the compiler assumes that behaviour.

[IntegratedQuantum]

Right, it should be noted that std.ascii.isPrint now also considers spaces to be printable. This is a bit of behavioral breaking change.

I'm not talking about the spaces.
Also the old version was considering spaces as printable:

pub fn isPrint(c: u8) bool {
    return inTable(c, tIndex.Graph) or c == ' ';
} //                                   ^ Spaces returned true in the old code

The problem is non-ascii characters(c >= 0x80).
In the old code the lookup table was filled with zeroes for non-ascii characters:
mem.set(u8, table[128..256], 0); (line 293)
So for non-ascii characters isPrint(c) == false == isControl(c) was true in the old code.
In the new code isPrint(c) == true for non-ascii characters.

And I think it makes sense to consider non-ascii characters as not printable. For example in UTF-8 all non-ascii characters consist of multiple bytes, so a single byte of that sequence cannot be printed.

Your changes prevent both src/DepTokenizer.zig and std.fmt.fmtSliceEscape* from escaping non-ascii characters.
std.fmt.fmtSliceEscapeLower is used in escapeUnprintables in src/translate_c.zig.
This would explain why test/behaviour/translate_c_macros_not_utf8.h is failing with invalid byte: '\xa9' on the CI.