Fix batch experiment calibrations

[coruscating]

Summary

Closes #631. Copies over calibrations when constructing batch experiment from children, like parallel experiments do.

[wshanks]

The basic change here looks good to me outside of the circuit.decompose() issue. I wonder if we could add a test for #631 as well to catch this kind of issue from coming back? Something that creates a dummy experiment with a calibration and checks that the circuits can be transposed when the experiment is used with a batch experiment?

[eggerdj]

Can we add a test for this? I.e. setup a batch experiment with e.g. two cal experiments and check that calibrations are there?

[coruscating]

This bug seems to not be due to calibrations, because it's been solved by #667. I will open another PR for adding tests.

[wshanks]

I think the issue might still be calibrations but #667 changed the way batch experiments' circuits are handled. Now the individual experiments are transpiled before the qubits are remapped into a new circuit. Before #667, the qubits were remapped before transpilation. My guess is that transpilation embeds the calibrations into the circuit so that they don't get lost in the remapping step, but I have not checked this.

[yaelbh]

1. If what @wshanks is saying is correct then perhaps @chriseclectic can remove these lines in his PR on parallels Transpile parallel experiment sub circuits #680.
2. We may still need these lines for the case of a call to circuits without transpilation.

When a bug gets unexpectedly fixed, it still remains in the status of a bug until we fully understand how it was fixed, and are convinced that there are no cases where the bug can still occur.