Fix MethodError while showing MethodError from invoke with UnionAll

[Seelengrab]

The old code had a comment about this being either a Tuple or a type, and chose to check for the type. Unwrapping the UnionAll and checking for being a tuple seemed to be the easiest way to fix this, so that's what I did.

Also adds the same comment from showerror to show_method_candidates, since that assumes the same thing.

Fixes #47559

[KristofferC]

Please use descriptive titles and commit messages.

[vtjnash]

Looks like a good start. Unfortunately, I have already worked with Aaron to write a more complete version of this PR (there are some additional subtleties to cleanup too around the use of the parameters). As a minor point of reference, for very new issues, you may want to explicitly confirm on Slack that it is not already being worked on, or at least assign yourself to it, to avoid repeating work.

This test may be helpful though, since we had not written that yet (@apaz-cli).

[Seelengrab]

Gotcha - the errorshow tests ran through locally, so I thought it should work (the test faillures seem to be unrelated - ReadOnlyMemoryError in SparseArrays?). I only looked at the issue in the first place since it had to do with showerror of MethodError, so I thought it was related to the printing changes. Feel free to close this if you've already worked out a better solution than my hacky unwrap_unionall :)

Assigning myself is not really an option without write access, but I'll ping on slack next time! 😅 The test more or less follows the pattern at the start of errorshow.jl, except with the sprint inside of the @test to capture that failure correctly.