Consider ADL-like MethodError warning

[Keno]

Consider

julia> module Foo
       struct Bar; end
       length(::Bar) = 1
       end
Main.Foo

julia> length(Foo.Bar())
ERROR: MethodError: no method matching length(::Main.Foo.Bar)
The function `length` exists, but no method is defined for this combination of argument types.

Closest candidates are:
  length(::Compiler.InstructionStream)
   @ Base ../usr/share/julia/Compiler/src/ssair/ir.jl:272
  length(::Compiler.DFSTree)
   @ Base ../usr/share/julia/Compiler/src/ssair/domtree.jl:121
  length(::Core.MethodTable)
   @ Base runtime_internals.jl:1746
  ...

Stacktrace:
 [1] top-level scope
   @ REPL[2]:1

The error here is the lack of import of Base.length. However, nothing in the error message hints you at that. We do have the reverse error message:

julia> Foo.length(1)
ERROR: MethodError: no method matching length(::Int64)
You may have intended to import Base.length
The function `length` exists, but no method is defined for this combination of argument types.

Closest candidates are:
  length(::Main.Foo.Bar)
   @ Main.Foo REPL[1]:3

Stacktrace:
 [1] top-level scope
   @ REPL[3]:1

I think it would be reasonable for the MethodError to look through all modules of types mentioned in the failed signature to see if they have a generic function matching the name of the failed generic function and hint that there may be a missing import.

[DilumAluthge]

Could we alternatively change the printing of the existing message to:

ERROR: MethodError: no method matching Base.length(::Main.Foo.Bar)

[Keno]

We could, but I don't think it adds anything in this case. Both of those are the correct generic functions and types. The issue is that Main.Foo.lengrh is different from Base.length by accident, but the user doesn't know that. I think an experienced Julia user would know to check, but might as well have the system do it for you.