Comparative benchmarks for hGetSome(at) and their replacement functions

[jorisdral]

A first step towards #50. These micro-benchmarks should allow us to determine whether we can replace the HasFS get-primitives by the HasBufFS primitives, without losing functionality or performance.

[jorisdral]

One thing I am unsure about is whether we can replace the HasFS put-primitives by the HasBufFS primitives easily. Some background first :

Defining hGetSome in terms of hGetBufSome requires conversion from a MutableByteArray to a ByteString, which is possible using unsafeByteArrayToByteString or copyByteArrayToByteString. The former function is a little bit hacky, but it ensures we don't copy bytes from the MutableByteArray to the ByteString, which is what the latter function would do.

Defining hPutSome in terms of hPutBufSome requires conversion from a ByteString to a MutableByteArray. This will not always be possible without copying, because a ByteString is a ForeignPtr under the hood which is a Ptr with optionally a reference to its underlying MutableByteArray. If the reference isn't there, then we can do nothing but copy the bytes to a fresh MutableByteArray.

The question is: do we want to keep both the "old" and the "new" primitives, if we can't ensure that the new primitives are as performant as the old ones?