clippy: fix raw ptr lints

[xtqqczze]

No description provided.

[clarfonthey]

The &raw const changes, I would say are an improvement, but I'm wary of the casts because hashbrown already has lots of places where we explicitly inline methods like this due to code bloat. Are you sure that isn't the case here?

I'd be in favour of consistency (right now, we have a mix of both), but it would make sense to move toward not doing this instead if that's the case.

[xtqqczze]

I haven’t seen any evidence of increased code bloat in the LLVM IR, when I checked on Godbolt (https://godbolt.org/z/PMKPfWWje) the generated code looked the same with and without cast.

[clarfonthey]

Right, the specific instances where code bloat was a concern was due to closures, which isn't a concern here. I think in that case it'd be reasonable to try and include some of the not-default but still useful pointer lints for clippy, to help ensure consistency.

[Amanieu]

In this case there is basically no bloat because the cast is inlined by the MIR inliner before monomorphization. The LLVM IR bloat issue really comes up with things like closure where inlining has to be deferred to post-monomorphization codegen.