Add support for double feature flags

[ThePumpingLemma]

While here, make some minor cleanups + reorganization for boolean
feature flags

[r3mariano]

I'm wary of potential floating-point precision losses with Double and the potential for confusion there, e.g. featureFlags.getDouble(...) != 0.3 despite it being set on the feature flag UI. (I see that the LD SDK uses a double too, and that uses Gson under the hood. See for instance google/gson#1127)

[escardin]

I'm wary of potential floating-point precision losses with Double and the potential for confusion there, e.g. featureFlags.getDouble(...) != 0.3 despite it being set on the feature flag UI. (I see that the LD SDK uses a double too, and that uses Gson under the hood. See for instance google/gson#1127)

Sure, but why do we care? Every flag I've looked at in older systems that does this is using it for something like a percentage rollout. The difference between 0.3 and 0.299999999999999988897769753748434595763683319091796875 for all practical purposes are the same. We also offer floats, and that's fine too.

I'm much more concerned about the double->long conversion, as I know people store epoch timestamps and expect them to parse, and that's too big to reliably represent

IMO if you need that level of precision in your decimal number... don't use a double, and you should already know that, because you learned not to use floating point for decimals/money already.

[r3mariano]

Sure, but why do we care? Every flag I've looked at in older systems that does this is using it for something like a percentage rollout. The difference between 0.3 and 0.299999999999999988897769753748434595763683319091796875 for all practical purposes are the same. We also offer floats, and that's fine too.

This isn't a percentage rollout, it's a decimal number flag value. I get that they are practically the same, but I'm not sure if a computer will always think so too, hence my previous example. In any case, happy to try it out.

IMO if you need that level of precision in your decimal number... don't use a double, and you should already know that, because you learned not to use floating point for decimals/money already.

Misk's a pretty "guard-rails"-heavy framework IMO. Even if parts of it do right now, we shouldn't rely on "you should know that".

[ThePumpingLemma]

@r3mariano are you okay with me merging this?