namespaced defines: -d:myproj.mykey to avoid clashes

[timotheecour]

defines have global scope and can cause clashes.
nim itself defines 363 unique defines, many of which are not prefixed by nim, see the list here: https://gist.github.com/timotheecour/8b2286478083a6068731a63114ee9dd0

The current scheme (which is rarely followed) is to prefix by nimble package name, eg: -d:nimHasLibFFI
However this scheme is brittle because of style and case insensitivity of defines, so you can still have clashes, eg:
-d:fooBarBaz # could be interpreted as (package foo, var barBaz) or (package fooBar, var baz) or any other variation

This is a problem because many nimble packages are prefixes of each other, eg see https://gist.github.com/timotheecour/182fce669ff217c6b0bc87d918303c3e which shows 233 (package1, package2) pairs such that package1 is a prefix of package2.
Furthermore, plenty of packages start with nim so that further causes chances of define clashes.

There is a simple solution to that:

proposal

allow using a namespace for defines, eg:

nim c -d:myproj.mykey1 main.nim
nim c -d:myproj.mykey2:foo main.nim

# main.nim
when defined(myproj.mykey1): discard
const bar {.strdefine:"myproj.mykey2".} = "foo2"

we can just restrict (for simplicity) to at most a single dot, so that foo.bar is valid but not foo.bar.baz

this allows a nimble package foo to prefix all its defines by foo., eg -d:foo.bar

implementation difficulty

easy, I can do it if this RFC is accepted

note

see also #269 for a proposal that enforces ODR (one definition rule), that is complimentary to this RFC.

[metagn]

Seems to be accepted?