fix #12130 ; improve naming scheme in fakePackageName

[timotheecour]

• 7 8 are more common (especially in tests) than @ so @ seems more appropriate mangling character
• should work on windows as had been mentioned in [TODO] Nim now allows modules with same name in a package #8614 (comment)
• should fix Filenames ending with xxx8.nim #12130 although I wasn't able to reproduce that issue on OSX (might affect linux only)
• /cc @qqtop @juancarlospaco does that fix the problem for you?

../bam@bax/cab.nim
=>
..@bam@@bax@cab.nim.c

../bam/bax/cab.nim
=>
..@bam@bax@cab.nim.c

the leading .. could be avoided but IMO is more readable and should be legal anyway

note

this PR does not try to make the fakePackageName a valid nim identifier (shouldn't be needed IIUC); note that this also wasn't the case before this PR, eg:
nimcache/_7bar.baz7gaz77.nim.c doesn't give a valid identifier (for ../bar.baz/gaz7.nim); ditto with starting with _7 or containing __)

[Araq]

You need to adapt some tests I think and then it's ready to go. In retrospect, I would have produced subdirs inside nimcache, but this improves the situation, so thank you!

[timotheecour]

You need to adapt some tests I think and then it's ready to go.

done

In retrospect, I would have produced subdirs inside nimcache, but this improves the situation, so thank you!

relative paths (..) is awkward to deal with subdirs though; and flat generated files seems easier to deal with (no need for rm -rf etc)

[dom96]

Thank you for this!

[juancarlospaco]

its not supposed to do this ❔
https://nim-lang.github.io/Nim/os.html#quoteShell%2Cstring
Because if not would be a great addition to that proc, like an optional escapeSlash = true or something.
🤔

[timotheecour]

@juancarlospaco not sure I understand how this relates to quoteShell; the goal of fakePackageName/demanglePackageName is to translate a relative path (possibly starting with ..) to a filename, and back. quoteShell has a different goal.