feat: relative weights in fractional, fix injected props

[bacherfl]

This PR adds support for relative weights in the fractional evaluator. This new (non breaking) feature has been proposed in open-feature/flagd#1282 (The related PR to implement this in flagd is this one: open-feature/flagd#1313).
In addition to supporting relative weights instead of percentages, the weight value for a distribution item can be omitted. In this case, a default weight of 1 will be applied.

Leaving this in draft until the related PR in flagd has been merged

[toddbaert]

@aepfli Not sure if .NET is your thing, but since you did a few others, could you quickly skim this?

[aepfli]

looks good from my side, i do have a little comment regarding the tests, but this is more like a little nit or improvement for the future

[aepfli]

we should evaluate if it makes sense to move the json tests of the java contrib sdk from https://github.com/open-feature/java-sdk-contrib/tree/main/providers/flagd/src/test/resources/fractional to the harness, and use them as test for all our implementation of the fractionals. maybe we are still missing some data, which is part of the test code, but i think this would allow us to be more stable throughout all the packages. Normally there should be a way in each of our contrib sdks to load json files and to parameterize tests in some sort of way (assumption for dotnet)

[toddbaert]

I found a couple small issues with this when adding some additional e2e tests, working on some fixes.

[toddbaert]

@bacherfl I made a couple minor tweaks to the algorithm; we had a slight deviation when no bucking value was supplied. You can see them here. Also, this was fun.

I also added the e2e tests for this new functionality.

Thanks so much!

[toddbaert]

This stumped me for a moment. In prod, from here implements IDictionary, but is not actually an instance of Dictionary, so this evaluated to false, resulting in no injected props. This fixes it!