Better interface for re-using routines

[mstechly]

Currently, the supposed workflow for using Routines is as follows:

1. Define a routine – both in terms of structure and resources.
2. Compile this routine.
3. Validate compiled routine. (Not necessary, but probably a good thing to do.)

At this point, we can say we have "validated" a given estimate. Let's say we did it for a QROM.

Since we might not wish to specify routine's substructure in future, higher-level routines (such as using a QROM as a black-box subroutine in Double Factorization), we would now want a new workflow for using product of the above one, i.e.:

1. Define routine for a higher-level algorithm
2. Import pre-defined, compiled and validated routines for any black-boxed subroutines.
3. Compile as before.
4. Validate as before.

Hence, we would therefore like a way to straightforwardly "access" a subroutine's precompiled, prevalidated resources and ports.

There are two things that should be done to enable this:

1. Pre-validated estimators should be accessed via a different mechanism than uncompiled ones. However, the user shouldn't have to care about whether the estimator they're using is really atomic, prevalidated, or anything else.
2. Compilation and validation of the routines that has been already compiled and validated

Originally by @sammorley-short