Make Complex observables evaluable

[hz-xiaxz]

related to (#12 )
After some searching, I found this package BinningAnalysis.jl implements likewise jackknife resampling method from QMC results, and he follows the julia convention to calculate var(x) = var(real(x))+var(imag(x)). I think this is reasonable enough then.

In this PR I made the type of error be alway :<Real and fixed some issue in reading the complex observable from the json result file.

[lukas-weber]

Awesome, thanks!

Apart from the comments on the code, could you also run JuliaFormatter.jl on it and sign your commit?

[lukas-weber]

Maybe error in line 40 should already be created as a real array, then we do not need to convert.

[hz-xiaxz]

thanks. I implement error = real.(zero(complete_eval)) here, is there better way?

[lukas-weber]

Can we integrate all of this into one dispatch of sanitize?

Like

sanitize(m::AbstractDict, e) = isnothing(e) ? missing : complex(m["re"] ± e, m["im"])

This should also already catch the vector case because sanitize is called with a dot below.

[hz-xiaxz]

Seems good but I got the error track

ERROR: ArgumentError: broadcasting over dictionaries and `NamedTuple`s is reserved
Stacktrace:
  [1] broadcastable(::Dict{String, Any})
    @ Base.Broadcast ./broadcast.jl:744
  [2] broadcasted(::Carlo.ResultTools.var"#sanitize#11", ::Dict{String, Any}, ::Float64)
    @ Base.Broadcast ./broadcast.jl:1344
  [3] measurement_from_obs(obsname::String, obs::Dict{String, Any})
    @ Carlo.ResultTools ~/.julia/dev/Carlo/src/resulttools/ResultTools.jl:23
  [4] (::Carlo.ResultTools.var"#3#5"{Dict{String, Any}})(obsname::String)
    @ Carlo.ResultTools ./essentials.jl:0
  [5] iterate
    @ ./generator.jl:47 [inlined]
  [6] _all(f::Base.var"#384#386", itr::Base.Generator{Vector{String}, Carlo.ResultTools.var"#3#5"{Dict{String, Any}}}, ::Colon)
    @ Base ./reduce.jl:1297
  [7] all
    @ ./reduce.jl:1283 [inlined]
  [8] Dict(kv::Base.Generator{Vector{String}, Carlo.ResultTools.var"#3#5"{Dict{String, Any}}})
    @ Base ./dict.jl:111
  [9] dataframe(result_json::String)
    @ Carlo.ResultTools ~/.julia/dev/Carlo/src/resulttools/ResultTools.jl:35

The broadcast will have an issue when m is a single dict. So I guess we should deal it first.

[hz-xiaxz]

Thanks for the review. As I'm new to Github, how can I sign for my commit?

[lukas-weber]

https://docs.github.com/authentication/managing-commit-signature-verification/about-commit-signature-verification

But, since signing existing commits can be a bit tricky, we can skip signatures for this one. :)

I also noticed CI is failing for nightly, but it seems unrelated to your changes.