fix(riscv): allow vfncvt FP-to-int narrowing at SEW=8

[carlosqwqqwq]

fix(riscv): allow vfncvt FP-to-int narrowing at SEW=8

Closes #211

Summary

The FP-to-integer vfncvt variants (vfncvt.xu.f.w/vfncvt.x.f.w/vfncvt.rtz.xu.f.w/vfncvt.rtz.x.f.w) convert a 2*SEW FP value to an SEW integer and are legal for every SEW: at SEW=8 a 16-bit FP source converts to an 8-bit integer. The previous SEW restriction (introduced by the #135 fix) rejected SEW=8 for all variants, which over-rejects the legal FP-to-int encodings. The int-to-FP and FP-to-FP variants still produce an SEW FP value and stay reserved at SEW=8.

Validation

• ftoi gains the 8-bit saturating paths (NaN -> i8/u8 max, out-of-range saturation, RNE rounding) used by the newly legal encodings.
• Regression verifies values against QEMU/native references: 1.5 -> 2, -1.5 -> -2, NaN -> 255 (u8), and that the int-to-FP variant at SEW=8 stays reserved.
• cargo test --lib: 8213 passed, 0 failed.

Notes

The vector semantics live in cpu/vector.rs after the refactor in #202, so this change does not grow the oversized cpu.rs.

[19h]

Thanks @carlosqwqqwq. The legal SEW=8 FP-to-integer narrowing fix is now consolidated into #203 in commit d1de115, with you credited as co-author. The rollup keeps FP8-producing variants reserved and adds exact rounding, saturation, NaN/infinity, NV/NX, SMIR interpreter, O0/O2 JIT, and live QEMU coverage. Closing this PR in favor of #203.