nir: CL size/align of bool (b1) should be 1, not 0

Without this, nir_lower_vars_to_explicit_types ends up producing
a scratch_size of 0 if there are any function_temp bools. I expect
they should get removed via optimizations later on, but it's a
simple fix to avoid having to try to prematurely optimize.
71 jobs for !6310 with upstream/bool-1-byte
merge request