[TASK]: Investigate MIRA MMT cache

[mwdchang]

Preamble

We current have a taskrunner task amr-to-mmt that converts AMR to MMT format, in a number of occasions this is doing redundant work, as we have, for example in stratify:

stratify(AMR, ...) => MMT_new => AMR_new

And we subsequently covert AMR_New to MMT_new again. This has been observed to be quite slow for large models - eg models with 300+ states. In models with 1000+ states this task could take 10s of seconds before completion.

This back and forth conversions is due to historically, AMR encompass more modelling frameworks than those represented in MMT, but we have since dropped support for these frameworks.