kvx: ubsan: integer overflow

This fixes a few places where ubsan complains about signed integer
overflow when running the testsuite, and that clz(0) is undefined.
When fixing the clz problem, I also noticed that we'd get complaints
if pval is ever LLONG_MIN.  Fix that by using unsigned arithmetic.

	* config/kvx-parse.c (get_token_class): Avoid signed overflow.
	Don't clz(0).
	* config/tc-kvx.c (PARALLEL_BIT): Avoid signed overflow.
2 files changed