[PATCH] Fix the failure exposed by checking constant integer argument range

[Tim Northover]

> The C test case changes can show this clearly. Previously, all the test case
> only pass constant 0 or 1 as the lane argument.

Thanks for the examples Jiangning. I'd fiddled with a couple of them,
but obviously happened to pick cases that already worked so got myself
confused.

> Here, if we don't remove the concat_vectors at lowering stage, we would have
> to explicitly describe concat_vectors in the pattern being matched.
> Otherwise, we would fail to combine fneg and fma to be a fms.

Ah, that makes sense. I think these look fine then.

Cheers.

Tim.