[0029] Cooperative Vectors: consider using unique interpretation enums for variable vs memory

[damyanp]

#398 (comment)

There are two types of interpretations in the spec currently:

1. Interpreting vectors stored in variables
2. Interpreting vectors / matrices store in memory

These both use the same enum. This means that there are values in the enum that don't apply to the memory resources (ie the packed ones) that we'll need to do work to prohibit using.

If we used a different enum for these then we woudn't have this problem. Having an additional enum might be confusing / annoying to work with though.