unicode/norm: simplify short source detection in Form.transform

Form.transform returns ErrShortSrc when not all source bytes
have been processed:

	if err == nil && n < rb.nsrc && !atEOF {
		err = transform.ErrShortSrc
	}

Simplify this check to check whether all bytes in src have been
processed (in which case the Transformer interface says we need
to return ErrShortSrc):

	if err == nil && nSrc < rb.nsrc {
		err = transform.ErrShortSrc
	}

This change has no actual practical effect.

Changing n (number of bytes handled in the current iteration)
to nSrc (total number of bytes handled) looks like a bugfix,
since we don't want to return ErrShortSrc just because
we ran for more than one iteration.

However, when atEOF==false, Form.transform never processes
its complete input (Because quickSpan won't process *its*
complete input except when at EOF, I believe because it's
waiting to see if any future characters combine with the
current input suffix.) It always returns ErrShortSrc
with or without this change.

Dropping !atEOF looks like it would cause us to return
ErrShortSrc when at EOF and not processing all source
characters (which would be correct behavior), but
Form.transform always processes the complete source when at EOF.

To summarize: This change is a complete no-op, but the
code is simpler, more obviously correct, and more robust
to future changes elsewhere. (For example, if something
causes Form.transform to fail to process the complete source
when at EOF, it will now correctly return ErrShortSrc.)

Change-Id: Id76f3dea328ac8407936adfcfb14b1066a6a6964
Reviewed-on: https://go-review.googlesource.com/c/text/+/794102
Reviewed-by: Neal Patel <nealpatel@google.com>
Reviewed-by: Neal Patel <neal@golang.org>
LUCI-TryBot-Result: golang-scoped@luci-project-accounts.iam.gserviceaccount.com <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Auto-Submit: Damien Neil <dneil@google.com>