Interface tests

[lorenzoh]

This adds interface tests to various interfaces in FastAI.jl. This allows

• giving user-friendly error messages by performing checks in high-level functions
• giving help when implementing an interface, and providing tests for implementations

Examples

Blocks

is_block(block, obs) checks that a piece of data is a valid observation for a Block:

When data is not valid, you get detailed information on the reason why:

This is extensible for custom Block types. Since tuples of blocks are also valid blocks, we can check them, too:

Data containers

is_data(data) checks that data implements MLUtils.jl's data container interface.

When that is not the case, precise errors are given indicating which parts are not implemented:

is_data(data, block) additionally checks that the observations in the data container are valid for a Block:

Status

This depends on Invariants.jl which still needs to be released.