This adds some simple sorting of constraints. There are a few potential
deficiencies:

1. It doesn't sort nested constraint tuples. We could solve this by
   carrying around some information about whether we are in a constraint
   context in `p_hsType`, but that seeemed moderately disruptive.
2. It doesn't handle constraint tuple synonyms. I think this is
   unavoidable since we can't know they're constraint tuples without type
   information.
3. The sorting is very crude, just sorting on the `Outputable`
   representation of  the type. I think this will more-or-less doe what
   people expect, but perhaps it's too expensive?

Fixes fourmolu#247

This adds some simple sorting of constraints. There are a few potential
deficiencies:

1. It doesn't sort nested constraint tuples. We could solve this by
   carrying around some information about whether we are in a constraint
   context in `p_hsType`, but that seeemed moderately disruptive.
2. It doesn't handle constraint tuple synonyms. I think this is
   unavoidable since we can't know they're constraint tuples without type
   information.
3. The sorting is very crude, just sorting on the `Outputable`
   representation of  the type. I think this will more-or-less doe what
   people expect, but perhaps it's too expensive?

Fixes fourmolu#247

This adds some simple sorting of constraints. There are a few potential
deficiencies:

1. It doesn't sort nested constraint tuples. We could solve this by
   carrying around some information about whether we are in a constraint
   context in `p_hsType`, but that seeemed moderately disruptive.
2. It doesn't handle constraint tuple synonyms. I think this is
   unavoidable since we can't know they're constraint tuples without type
   information.
3. The sorting is very crude, just sorting on the `Outputable`
   representation of  the type. I think this will more-or-less doe what
   people expect, but perhaps it's too expensive?

Fixes fourmolu#247

This adds some simple sorting of constraints. There are a few potential
deficiencies:

1. It doesn't sort nested constraint tuples. We could solve this by
   carrying around some information about whether we are in a constraint
   context in `p_hsType`, but that seeemed moderately disruptive.
2. It doesn't handle constraint tuple synonyms. I think this is
   unavoidable since we can't know they're constraint tuples without type
   information.
3. The sorting is very crude, just sorting on the `Outputable`
   representation of  the type. I think this will more-or-less doe what
   people expect, but perhaps it's too expensive?

Fixes fourmolu#247

This adds some simple sorting of constraints. There are a few potential
deficiencies:

1. It doesn't sort nested constraint tuples. We could solve this by
   carrying around some information about whether we are in a constraint
   context in `p_hsType`, but that seeemed moderately disruptive.
2. It doesn't handle constraint tuple synonyms. I think this is
   unavoidable since we can't know they're constraint tuples without type
   information.
3. The sorting is very crude, just sorting on the `Outputable`
   representation of  the type. I think this will more-or-less doe what
   people expect, but perhaps it's too expensive?

Fixes #247