Update spec to define "strong reproducibility" and "weak reproducibility"

[alilleybrinker]

This is something @AevaOnline has raised in several past discussions, and which interrelates with our ongoing discussions on convergence and divergence (#42) and representation of build tool data (#56).

Currently "strong reproducibility" accords with the definition of "reproducible" in the spec, namely:

Reproducible: Independent parties, presented with equivalent artifacts, derive the same artifact identity.

This is byte-for-byte equality, and doesn't account for the possible need to strip additional data out (as would be required for "weak reproducibility" as I understand it.