implementing 0-based indexing

Hello all. I want to implement some StaticArray objects (LorentzTensor) with 0-based indexing and some custom methods. (I'm on 0.7.) I made the appropriate definition of axes and getindex(t::LorentzTensor, i::Int). This works fine for vectors, but (as you probably know) fails for all higher rank tensors. I've traced the problem down to here which indeed seems to be specialized to 1-based indexing.

My question is, what is the minimum I should implement to get this working? Will there be some other nasty surprises?

I would be happy to make a PR using axes to get this working for the general case, but I'm assuming that it would be far easier said than done (I have no idea what else it would break) and considering that this is a performance-critical package I don't know what other side effects this would have.

Start by reading src/indexing.jl at line 22 and compare its indexing assumptions with the issue's axes and getindex definitions for LorentzTensor. Investigate how higher-rank tensors are handled and identify the tests needed to establish correct 0-based indexing without affecting existing StaticArray behavior. Done means the supported implementation scope and regression coverage are clear.