Storage: remove dead-code (and clarify the encoding of filenames)

Following @raphael-proust reviews in !2466 (merged) I've tried to make sure the V1 functions are not accessed by mistake (they should only be used to compute preimage for hashes). This actually lead me to the conclusion that some of that code was wrong and was working only by luck.

This MR simplifies the offending code and explain why it actually works. It also adds a few tests to make sure the changes do not actually change the context hash.

Note: this shows that the encoding for filenames longer than 128 characters is wrong in tezedge and that we need to document the context hash properly (we have started to do so anyway). This doesn't have any impact right now as 1/ users cannot create arbitrary filenames and 2/ the existing filenames all have a length smaller than 127.