Flag format: uint8 vs bool

[ludwigschwardt]

From the MSv4 schema it seems that the FLAG data variable is of type bool. Looking at Zarr, this is mapped to a single byte (not a bit), although the v3 spec leaves the door open:

Zarr v3 is limited to type sizes that are a multiple of 8 bits but may support other type sizes in later versions of this specification.

The reason this interests me is because MeerKAT's MVF4 format uses uint8 for flags and actually store 8 different flags in there. It would be a shame to expand this to 6 separate flag bytes (we have 2 unused bits at the moment), although I suspect compression might come to the rescue. This is in light of the absence of flag categories in the new format.

My current thinking is to pretend that our uint8s are bools by saving our flags as flags.view(bool).

Would this be "unsafe" behaviour? What about exploring the idea to make the flags uint8 from the outset and support categories that way?