Code actions for replacing wildcards in types broken in GHC 9.8

[jhrcek]

There's a regression in GHC 9.8 in how types of wildcards in type signatures are inferred.
We added a conditional in tests to adjust to this new behavior - see comment.

I reported it in ghc and it got rejected as "this is the new behavior now, we should just document it"

I don't think we can easily change back the behaviour (in GHC 9.6 we defaulted the type variable, which we no longer do in 9.8).
See ghc issues for details.
https://gitlab.haskell.org/ghc/ghc/-/issues/24522
https://gitlab.haskell.org/ghc/ghc/-/issues/24425

This change in behavior breaks 2 tests:

haskell-language-server/plugins/hls-refactor-plugin/test/Main.hs

Lines 665 to 673 in 9076f90

	, testUseTypeSignature "multi-line message 1"
	[ "func :: _"
	, "func x y = x + y"
	]
	[ if ghcVersion >= GHC98
	then "func :: a -> a -> a" -- 9.8 has a different suggestion
	else "func :: Integer -> Integer -> Integer"
	, "func x y = x + y"
	]

haskell-language-server/plugins/hls-refactor-plugin/test/Main.hs

Lines 695 to 703 in 9076f90

	, testUseTypeSignature "no spaces around '::'"
	[ "func::_"
	, "func x y = x + y"
	]
	[ if ghcVersion >= GHC98
	then "func::a -> a -> a" -- 9.8 has a different suggestion
	else "func::Integer -> Integer -> Integer"
	, "func x y = x + y"
	]

So I think we should just update the test not to rely on polymorphic type and type defaulting
e.g.

func :: _
func x y = x && y

[jhrcek]

It’s not actually broken, just works differently now (no type defaulting). Added comment about it in #4138. Not much more we can do about it now.