Standardise Loss Reduction Modes

[nmichlo]

Removed reduction="sum" in #12, but considerable effort is still required to maintain and verify models (#3, #4, #5)

Two options exist:

1. Allow models to specify their allowed modes: "mean" or "mean_sum"
2. Only use "mean_sum"

[nmichlo]

"Loss reduction mode" should maybe be called "beta scaling" and be a boolean value?

• This takes inspiration from the original BetaVAE paper if I am not mistaken.

The problem with the loss reduction modes is that not all the frameworks supports them.

• Could this automatically be handled by changing how the loss is aggregated?