Refactor/core

[leissa]

Changes

• %core.nop -> %core.nat: nop usually means sth entirely differently.
• Use implicit Dst type for %core.bitcast: .ax %core.bitcast: Π [D: *, S: *][S] -> D -> .ax %core.bitcast: Π.[S: *][D: *][S] -> D
• showcasing some of the newer features of Thorin in lit/clos/malloc.thorin
• adds wrap around parameter to deal with non-power of two .Idx in core::bit1 & core::bit2 See Discord discussion. Instead of casting to bit_floor or adding proof certificates, I just added the solution we already have for %core::wrap: add another (.Nat) flag that determines the overflow behavior. Rn, however, nobody is really doing anything with this flag. This is sth for future work.

[NeuralCoder3]

Use implicit Dst type for %core.bitcast:

The source type is now implicit not the destination type, right? (The dst type could theoretically be implicit but could not be locally inferred; the information is only available in its usage context (Hindley–Milner style inference/unification))

[leissa]

Use implicit Dst type for %core.bitcast:

The source type is now implicit not the destination type, right? (The dst type could theoretically be implicit but could not be locally inferred; the information is only available in its usage context (Hindley–Milner style inference/unification))

correct.