Eval plugin: proper multilined results handling and command-name abbreviations

[konn]

This pull-request consists of two changes to Eval plugin:

1. Bugfix: a proper treatment of multilined results.

   Previously, GHCi-like commands in Eval plugin doesn't treat multilined results properly.

   For example, with the current Eval plugin we get:

   f :: forall k n a. (KnownNat k, KnownNat n, Typeable a)
     => Proxy k -> Proxy n -> Proxy a -> ()
   f _ _ _ = ()
   
   -- >>> :type f
   -- f :: forall k1 (k2 :: Nat) (n :: Nat) (a :: k1).
   (KnownNat k2, KnownNat n, Typeable a) =>
   Proxy k2 -> Proxy n -> Proxy a -> ()

   Apparlently, the last two lines must be prefixed with --.
   This pull request addresses this issue and now it seems that the plugin returns the same result as GHCi:

   -- >>> :type f
   -- f :: forall k1 (k2 :: Nat) (n :: Nat) (a :: k1).
   --      (KnownNat k2, KnownNat n, Typeable a) =>
   --      Proxy k2 -> Proxy n -> Proxy a -> ()

   It tries to use Outputable.SDoc to generate the same alignment as the GHCi.

2. New feature: abbreviation for GHCi commands.

   Eval plugin now lookups GHCi-like commands first by exact match, then fallbacks to the first element in the dictionary with the given prefix.
   Since there are just three commands currently, I just used lists to implement command dictionary; perhaps we should use some kind of trie in a future?

[Ailrun]

Minor style comments

[konn]

I've forgotten to fix the pragma styles in functional tests. Now they must be accord with the existing codebase's LANGUAGE pragma style.

[pepeiborra]

Thank you!