REPL: enable MethodCompletions after keyword arguments

[aviatesk]

With this PR, we can have completions for e.g. foo(1, kw1 = 1, ...) when we have foo(a; kw1 = 1, kws...) = ...:

julia> foo(a; kw1 = 1, kws...) = return
foo (generic function with 1 method)

julia> foo(a::Integer; kw1 = 1, kws...) = return
foo (generic function with 2 methods)

julia> foo(1, )

foo(a::Integer; kw1, kws...) in Main at none:1 foo(a; kw1, kws...) in Main at none:1
julia> foo(1, kw1 = nothing, )

foo(a::Integer; kw1, kws...) in Main at none:1 foo(a; kw1, kws...) in Main at none:1
julia> foo(1, kw1 = nothing, kw = '1', )

foo(a::Integer; kw1, kws...) in Main at none:1 foo(a; kw1, kws...) in Main at none:1

It will also benefit certain frontends like Juno (and possibly julia-vscode if we merge this PR), allowing them to predicate return type on completion signatures from keyword method.

[aviatesk]

Without this, the completion would look like below and imho looks less friendly (while it's the internal truth):

julia> foo(1, kw1 = nothing, )

(::var"#foo##kw")(::Any, ::typeof(foo), a::Integer) in Main at none:1 (::var"#foo##kw")(::Any, ::typeof(foo), a) in Main at none:1

q: is there a way to retrieve the original Method object from its keyword-arg Method ? Then we don't need this additional field.

[fredrikekre]

Fixes #37607?

[aviatesk]

yep

[aviatesk]

@JeffBezanson could you have a review on this ?

[aviatesk]

Could someone review this ? Maybe @JeffBezanson @pfitzseb or @KristofferC ?

[KristofferC]

Tests look ok and it's just an interactive feature so I'll just merge this after CI is green.

[aviatesk]

Thanks, I just added some type annotations to improve type stabilities around complete_methods

[aviatesk]

The failures in CI are unrelated:

    REPL                            |     1223                    3      1226
    Profile                         |       22     1      1                24
    Serialization                   |      119                            119

[aviatesk]

@KristofferC bump.