Remove dead magics

[krux02]

I went through all the magics and verified if they are still used. If not all their code is deleted.

Note:

I deleted mSymDiffSet as it seems to be not used anymore. Is that a bug?

[krux02]

test failure unrelated. It is yet again "https://github.com/status-im/nim-blscurve"

[timotheecour]

how about using #11424 all the way, (ie replacing each comment by a litteral index as you started doing) => would guarantee it does the right thing at CT, especially as some magics are being removed

but this also could be done in later PR's

[krux02]

I already did it in my other PR: https://github.com/nim-lang/Nim/pull/13536/files

[timotheecour]

we don't need 2 ways of doing the same thing but maybe the non string way is better (because simpler), ie some (future cleanup PR) could change these magics to use the non-string variant everywhere or even use a similar trick as importc and omit the name altogether when its derivable from the symbol name eg:

proc foo() {.magic: "Foo".} => `proc foo() {.magic.}`

[Araq]

The string version doesn't universally work, for string* {.magic: "string".} the compiler used to complain no string type was available for "string".

[timotheecour]

The string version doesn't universally work

that doesn't seem incompatible with what i'm suggesting; Im suggesting:

• change all string version to identifier versions
• and optinally (like for importc), for the common case where identifier is predictable from symbol name (modulo capitalizeFirstLetter), skip it

type string* {.magic: String.} is already in non-string mode; and, under optional rule, would become type string* {.magic.} since "String"==capitalizeFirstLetter("string") ; note that the magic enum is still mString, that's not changing

proc `or`*(a, b: typedesc): typedesc {.magic: "TypeTrait", noSideEffect.}

would become (since optional shortcut doesn't apply):

proc `or`*(a, b: typedesc): typedesc {.magic: TypeTrait, noSideEffect.}