Migrate execution to dynamic dispatch

[ivanlele]

This PR closes #349 by moving all non-object-safe Jet supertraits onto Box and migrating the project’s internal generic model to dynamic dispatch.

[ivanlele]

@apoelstra The main voodoo lives in stc/bit_machine/mod.rs:

let typed_jet = jet
    .as_ref()
    .as_any()
    .downcast_ref::<JE::Jet>()
    .ok_or(ExecutionError::JetTypeMismatch)?;
jet_result = self.exec_jet(typed_jet, env);

Here we downcast the dyn Jet into a concrete type so it can be matched inside c_jet_ptr.

With the move to dynamic dispatch, we lose some benefits, a few places now require additional generic boilerplate (prime example is: RedeemNode::decode::<_, _, Core>). But, we also received some: working with plain Node types becomes simpler since they no longer carry generics.

In a follow up PR, I'm thinking to add runtime checks for dyn Jet within Context, to somehow compensate for the loss of compile time checks (runtime performance will take a hit unfortunatly)

[apoelstra]

At a conference and have only done a cursory review so far. Overall this looks great! Amazing to get rid of so many generics. Frustrating to have extra generics on Redeem::decode but we expected this (and conceptually I think it's necessary to do it this way, or an equivalent way like taking a decode_jet closure).

@apoelstra The main voodoo lives in stc/bit_machine/mod.rs:

Neat. We can revisit this in a followup. Maybe we can remove the generic from BitMachine somehow by doing dangerous things with c jet ptrs? Unsure. Would likely require more changes on the Haskell side. Let's definitely not try to do it in this PR, which is already super big, so for now this solution is good.

In a follow up PR, I'm thinking to add runtime checks for dyn Jet within Context

Yeah, this is a good idea (but should be in a followup). We may want a JetType enum or something which can be returned from the DynJet trai. Unsure.

[apoelstra]

ACK 0fb7c51; successfully ran local tests