Parallelize the pipeline

[janmasrovira]

As part of the effort to improve the efficiency of the compiler and improve tooling responsiveness (#2685), we should parallelize the pipeline.

The idea is to build an import tree where the vertices are juvix (top) modules and the edges represent import dependencies.
Then, we create a compile queue from which worker threads pull pending modules and compile them in parallel.
A toy experiment that simulates this strategy has been implemented successfully in https://github.com/anoma/juvix/blob/parallel/app/Parallel/Experiment.hs

How much of the pipeline will be parallelized is yet to be decided. My initial goal is to parallelize (at least) the deserialization of the .jvo files. After that is done, we can extend it further.

Naturally, the efficiency gains from this will depend on the relation between the height and the size of the import tree. My rought expectation is that it will be at least 3x faster on average.

[janmasrovira]

Closed by #2779