[stdlib] Elide some null-pointer checks in UM[R]BP bulk-copy functions

[glessard]

Writing these functions with calls through optional chaining (the ? operator) meant that an extra nil check could never be elided from code that uses them. We replace the ? operator with unsafelyUnwrapped instead, providing safety in debug mode in the presence of a malformed UnsafeMutableBufferPointer, while eliding the nil check in release mode.

[glessard]

@swift-ci please test

[stephentyrone]

Looks good, thanks.

I do have a question of whether we need to no-op these out when source.isEmpty or we can allow that case to fall through, which I think boils down to a question about the semantics of Builtin.assignCopyArrayFrontToBack and friends. I asked Andy and Arnold to weigh in on that on slack.

There's one comment that implies that it's OK:

// This builtin is equivalent to:
// for i in 0..<count {
//   self[i] = source[i]
// }

but I'm not certain that this comment intended to document the behavior in the count == 0 case.

[stephentyrone]

Ok, clarified that count == 0 is OK, here, please merge.