Consider letting x... mean (x...,)

[MasonProtter]

Currently this errors:

julia> (2,3)...
ERROR: syntax: "..." expression outside call around REPL[75]:1

but this doesn't:

julia> 1, (2,3)...
(1, 2, 3)

Obviously, letting (2, 3)... == (2, 3) doesn't seem particularly useful on the surface, but I do find it useful in macros where I would like f(@g x, y) to mean f(@g(a), b) and @g x, y to mean (@g(x), y) which currently errors if @g(x, y) returns an Expr(:...). The reason I need to write the macro this way is because of #36547.

See for example #38713 (comment) and https://github.com/MasonProtter/SimpleUnderscores.jl#caveats