Claude review: gpu: nova-core: gsp: move chipset-specific parts of the boot process into a HAL

[Claude Code Review Bot <claude-review@example.com>]

Patch Review

The HAL design is clean:

```rust
pub(super) trait GspHal: Send {
    fn boot(&self, ...) -> Result;
    fn post_boot(&self, ...) -> Result { Ok(()) }  // default impl
}
```

The `post_boot` default implementation returning `Ok(())` is good — Hopper/Blackwell won't need the GSP sequencer.

**`gsp_hal()` dispatch**:
```rust
pub(super) fn gsp_hal(chipset: Chipset) -> &'static dyn GspHal {
    match chipset.arch() {
        Architecture::Turing | Architecture::Ampere | Architecture::Ada => tu102::TU102_HAL,
        Architecture::Hopper | Architecture::BlackwellGB10x | Architecture::BlackwellGB20x => {
            gh100::GH100_HAL
        }
    }
}
```

Using `&'static dyn GspHal` with const instances is a nice pattern that avoids runtime allocation while still being extensible.

**GH100 stub**: Returns `Err(ENOTSUPP)` which maintains the previous behavior for unsupported architectures. Good.

**`boot()` signature**: `#[allow(clippy::too_many_arguments)]` — 8 arguments is a lot, but given the hardware initialization context, all parameters are genuinely needed. Could be packed into a struct but that would be churn with no real benefit.

**Pin<&mut Gsp> change**: `self: Pin<&mut Self>` → `mut self: Pin<&mut Self>` — needed because `self.as_mut()` is called to reborrow the pinned reference. Correct.

No issues.

---

---
Generated by Claude Code Patch Reviewer