[llvm] [RISCV][RegAlloc] Add getCSRFirstUseCost for RISC-V (PR #131349)

[Luke Lau via llvm-commits]

lukel97 wrote:

> Also I don't see the big improvement on 523.xalancbmk_r on SiFive core, which makes me hesitate about the value 64.

>From Michael's results there's a 0.2% dyn instcount improvement and a 0.4% improvement on the BPI-F3, which seems to check out? Inspecting the codegen diff on -O3 -march=rva22u64_v there's lots of functions with diffs like e.g:

```diff
--- build.rva22u64_v-O3.a//External/SPEC/CINT2017rate/523.xalancbmk_r/CMakeFiles/523.xalancbmk_r.dir/Users/luke/Developer/cpu2017/benchspec/CPU/523.xalancbmk_r/src/Abstrac
tNumericValidator.s     2025-03-19 17:29:10
+++ build.rva22u64_v-O3.b//External/SPEC/CINT2017rate/523.xalancbmk_r/CMakeFiles/523.xalancbmk_r.dir/Users/luke/Developer/cpu2017/benchspec/CPU/523.xalancbmk_r/src/Abstrac
tNumericValidator.s     2025-03-19 17:27:46
@@ -79,27 +79,25 @@
        .cfi_personality 155, DW.ref.__gxx_personality_v0
        .cfi_lsda 27, .Lexception0
 # %bb.0:                                # %entry
-       addi    sp, sp, -64
-       .cfi_def_cfa_offset 64
-       sd      ra, 56(sp)                      # 8-byte Folded Spill
-       sd      s0, 48(sp)                      # 8-byte Folded Spill
-       sd      s1, 40(sp)                      # 8-byte Folded Spill
-       sd      s2, 32(sp)                      # 8-byte Folded Spill
-       sd      s3, 24(sp)                      # 8-byte Folded Spill
-       sd      s4, 16(sp)                      # 8-byte Folded Spill
+       addi    sp, sp, -48
+       .cfi_def_cfa_offset 48
+       sd      ra, 40(sp)                      # 8-byte Folded Spill
+       sd      s0, 32(sp)                      # 8-byte Folded Spill
+       sd      s1, 24(sp)                      # 8-byte Folded Spill
+       sd      s2, 16(sp)                      # 8-byte Folded Spill
+       sd      s3, 8(sp)                       # 8-byte Folded Spill
        .cfi_offset ra, -8
        .cfi_offset s0, -16
        .cfi_offset s1, -24
        .cfi_offset s2, -32
        .cfi_offset s3, -40
-       .cfi_offset s4, -48
        .cfi_remember_state
```

Given that the 

https://github.com/llvm/llvm-project/pull/131349