Generalize Buffer.BlockCopy optimization for either src or dst

[jamesqo]

Don't have that much experience in native programming, but here it goes...

I noticed from #3118 that an optimization was added in the case where both src and dst were byte arrays, but not when only one of them was. Therefore, the optimization would not apply to code like this:

var bytes = new byte[4] { 1, 2, 3, 4 };
var shorts = new shorts[2];
Buffer.BlockCopy(bytes, 0, shorts, 0, bytes.Length);

This PR changes that so the optimization runs if either the src or dst arrays is a byte array, otherwise falling back to the default implementation as usual.

cc @jkotas

[jkotas]

I have never seen Buffer.BlockCopy usage pattern like this. Do you have any examples of code where this helps?

Checking for a special case - that pretty much never happens - is making the code slower for everybody else. It is net loss - not making any real world workload faster.

[jamesqo]

@jkotas Sure, it's one of the top answers on Buffer.BlockCopy on StackOverflow: http://stackoverflow.com/a/3783189/4077294

Besides (correct me if I'm wrong) I don't think this is adding any overhead to the cases where this optimization doesn't apply.

[jkotas]

Typo: fro -> from

[jkotas]

I see. LGTM. Thanks!