FileCheck regex should support common regex escape sequences

[cjdb]

clang\d is a pattern that's recognised by many regex engines to mean clang[0-9], but FileCheck doesn't seem to recognise it. It would be good to have FileCheck recognise the following patterns:

• \f, \n, \r, \t, \v: usual escape sequences

• \b: matches a word boundary

• \d: equivalent to [0-9]

• \s: equivalent to [ \f\n\r\t\v]

• \w: equivalent to [A-Za-z0-9_]

• \B: inverse of \b

• \S: inverse of \s

• \D: inverse of \d

• \W: inverse of \w

The above are good for matching ASCII characters, but don't scale for anything that's outside of ASCII. If we're to add this feature, I think it would be good to produce a design that incorporates Unicode code points as well.

[asl]

The regex implementation available in lib/Support seems to support only POSIX-style regex'es. So, one could use [:digit:] instead of \d

Here is list of supported classes: https://github.com/llvm/llvm-project/blob/main/llvm/lib/Support/regcomp.c#L58

[cjdb]

[:digit:] is longer than [0-9] which already interrupts readability when compared with \d. Further, these types of escapes are accepted by a large variety of regex engines, and it was surprising to learn that FileCheck doesn't support this (I spent a couple of hours debugging before swapping out \d with [0-9]).

If POSIX regex doesn't support this, then we should consider expanding to a style that supports both [:digit:] and \d.