panfrost: Use the correct size for UBO loads

Multiply by the destination bit size to get the number of bits to load instead of assuming 32 bits.

Fixes flakes in many CL tests.

Fixes: 2e57684d ("pan/bi: Implement load_ubo with the builder")

<32-bit loads are currently broken. I guess the options for this MR are:

1. Ignore the problem and hope nothing loads less than 32 bits
2. Use MAX2(32, nir_dest_bit_size(instr->dest)) for now
3. Add my patches for fixing loads/stores of smaller sizes to this MR