* Remove unused Array.traverse{ST,IO} functions

* Remove unused Array.copyM function

* Remove unused Array.run2 function

* Remove unused Int argument from array function

* Remove copyMutableArray#, clean up imports

* Remove Int parameter from marray function

* Remove Array.lengthM and sizeOfSmallMutableArray#

* Revert "Remove unused Array.traverse{ST,IO} functions"

This reverts commit 492b394.

* Revert "Remove Array.lengthM and sizeOfSmallMutableArray#"

This reverts commit 3b04d96.

* Revert "Remove copyMutableArray#, clean up imports"

This reverts commit 126a7d5.

* Silence warnings regarding copyMutableArray#

* Remove [m]array smart constructors

* Revert "Remove unused Array.copyM function"

This reverts commit 089e9e4.

The latest ghc-head version is 8.7.20181217, so this job is quite
useless now.

See https://launchpad.net/~hvr/+archive/ubuntu/ghc

With GHC < 8.2, the standard `unsafeShift{L,R}` definitions
aren't inlined properly in larger unfoldings. This results
in benchmark slowdowns on the order of 1 to 6%.

- More information in introduction (basic operations, and using HashSet with custom
data types)
- Examples alongside function docs

This also removes some "Stability" annotations from internal modules.

Context: #211.

Fixes #287.

* add fast subset operation

* add explanation for subkey offset in lookupCont

* rename subset operation for compatibility with containers API

* update docs: union is not a least upper bound operator for `isSubmapOf`.

* explain runtime complexity of isSubmapOf.

* isSubmapOfBy: move `Empty` case to top

* isSubmapOfBy: fix comments

* isSubsetOf: add example

* isSubmapOf: quickcheck test for compatibility with containers

* isSubmapOf: use arbitrary instance of HashMap

* isSubmapOf: fix comments again

* isSubmapOf: update doc for runtime complexity

* remove mathematical symbols from user doc

* add difference subset quickcheck property

* add `all` function for arrays

* fix comments in `isSubmapOf`

* fix wrong runtime complexity of set inclusion

* delete unused property

* fix error in `isSubmapOf` based on wrong assumption

* add benchmarks

* change a few recursive `isSubmap` cases to `False`

* add strictness annotations

* make isSubmapOf and isSubmapOfBy INLINABLE

…to ensure that hyperlinks preferably target the public modules.

…not much use having them in PRs.