Move write cache to from datasaver to dataset

[jenshnielsen]

First step in creating a more clean interface between the dataset and the datasaver

[codecov]

Codecov Report

Merging #2112 into master will increase coverage by 0.03%.
The diff coverage is 88.07%.

@@            Coverage Diff             @@
##           master    #2112      +/-   ##
==========================================
+ Coverage   71.20%   71.23%   +0.03%     
==========================================
  Files         151      151              
  Lines       20148    20148              
==========================================
+ Hits        14346    14353       +7     
+ Misses       5802     5795       -7     

[astafan8]

Long time ago i made a mental model for myself for justifying the existence of DataSaver - it's that this object bridges DataSet that knows nothing about qcodes Parameters with the "measurement world" where poeple operate with qcodes parameters, including array parameters, multiparametersd, etc. So, with all that finalize_res_dict* stuff being moved here, which is done for the sake of DataSaver NOT having its "_results" cache, the point of DataSaver is largely gone.

So, my proposal would be to keep the finalize_res_dict* in DataSaver, and move only the "_results" part to dataset with the write_period and the flush_to_db method.

What do you think? Do you agree with the mental model?

as a side note - we can probably indeed move in the direction of getting rid of DataSaver but that's an API change, and i don't know if we'll do it anytime soon, so this is why i'm still folowing that mental model so that DataSaver has at least some justification for its existence :)

Pelase ignore this, i didn't read the PR well when i wrote it, sorry for that :)

[astafan8]

Approve! modulo guarding _enqueue_results with a check for DataSet state

[astafan8]

shall we keep this private? my thinking is if it's private now, then if we decide to change/remove/move it, we don't need to do deprecation cycles and stuff.