feat(developer): build caching framework for remote resources in keyboards and lexical-models

Currently we have a situation where a remote resource for the lexical-models repository is unavailable, which blocks the release build for that repository. We need to cache remote resources at submission time. This needs design, documentation, and development in a short timeframe.

Note that this should also be in consideration for epic/single-keyboard-repos -- we will need to implement a similar caching strategy there.