Faster drop of `steps`

[arthurpaulino]

In the implementation of prove_recursively for our Nova and SuperNova proofs, we take ownership of a vector of multiframes steps: Vec<C1LEM<'a, F, C>>. It's consumed and then dropped.

Here, it seems that we should be able to use the following trick: https://abrams.cc/rust-dropping-things-in-another-thread (thanks @huitseeker for the tip)

Note: we actually have two logical branches. In one branch, steps is consumed to instantiate cc. Meaning that on that branch specifically, cc is the thing we'd drop on a separate thread.