v3d: Implement a small immediates optimization, based on VC4's.

We can do one per instruction, and we have to be careful not to overwrite
raddr_b, but this greatly reduces the pressure on uniform loads
(particularly around ldvpm/stvpm instructions).

total instructions in shared programs: 90768 -> 88220 (-2.81%)
instructions in affected programs:     82711 -> 80163 (-3.08%)

src/broadcom/compiler/v3d_compiler.h

@@ -115,6 +115,7 @@ static inline struct qreg vir_reg(enum qfile file, uint32_t index)
  */
 struct qpu_reg {
         bool magic;
+        bool smimm;
         int index;
 };