add support for binaryOps for native histograms

[Saumya40-codes]

Found via #525

Prometheus supports the binary OR, AND, UNLESS operation when a native histogram is there in the query, this PR updates the related functions to take in account for native histograms.

[MichaHoffmann]

What about the other many to many joins "and"and "unless"

[Saumya40-codes]

What about the other many to many joins "and"and "unless"

Yes, "unless" needs to be implemented as well. Though, the implementation of end exists (and with current "or" implementation), ig it needs to handle few cases !??

• It works right when when both lhs and rhs are based upon only "non-native-histogram" or only "native histogram"

but in the below query:

histogram_quantile(0.6, request_duration_seconds_native_hist) and request_duration_seconds_native_hist 

o/p from thanos engine will be empty based on how we are handling here

but as labels of lhs and rhs matches, the o/p should be the answer of histogram_quantile(0.6, request_duration_seconds_native) (which is what coming up from prometheus engine)

---

such queries is redundant though 😅 no idea if this is something which should be changed or not though

[MichaHoffmann]

What happens if native histogram and float series collide in same hash bucket? Can you add a test for that?

We should behave same as Prometheus here even if the query might look weird

[Saumya40-codes]

What happens if native histogram and float series collide in same hash bucket? Can you add a test for that?

We should behave same as Prometheus here even if the query might look weird

Yup, made same changes where withMixedTypes = true is also checked against the three joins

[MichaHoffmann]

Can you please add some testcases for mixed (histogram vs float) many-to-many matches just to be super sure that those work as intended? Code itself looks solid to me and from running through it in my head i didnt see anything wrong so i would think those test cases will jsut work - but we should have them regardless

[Saumya40-codes]

Can you please add some testcases for mixed (histogram vs float) many-to-many matches just to be super sure that those work as intended? Code itself looks solid to me and from running through it in my head i didnt see anything wrong so i would think those test cases will jsut work - but we should have them regardless

I've added few cases where lhs and rhs are can be (native histogram or a float series with different/no-labels), does it needs any changes?