Introducing a new generic-solver stage?

[Eliyaan]

We recently introduced a new stage comptime for solving compile-time constructs like $if $match... .
I was trying to move more code from cgen comptime_if() to comptime but it required generics to be solved/solvable (because of gen_branch_context_string).
The question is do we solve (once and for all later stages) generics in comptime or do we introduce a new stage for this before comptime? (the new stage would replace the generic AST nodes by new solved ones, cloning the AST for each concrete type, see #26068)
I am in favor of introducing a new stage before comptime so each stage only does one job. Does solving generics have a dependancy on solving comptime constructs?

What do you think of introducing a new generic-solver stage?
@kbkpbot @felipensp