fixes, debugging, wip on DERelative tests

[dehann]

One issue remains with solveTree breaking the result. See broken_tests in this PR. The issue is that although initAll! does good job, the follow-on solve tree with .useMsgLikelihoods=false (over 4 cliques) resets some of the variables to values [0,0], and so:

• This problem might not be specific to DERelative, and will resolve in a next PR.
• Also going to use a separate PR to move sampling code from sampleFactor to getSample instead (finally able with new CalcFactor design).

some improvements relating to:

• restore alternate tests from #1716 / refactor #1721
• restore different point count N on solve graph, also fix test #1722

[codecov]

Codecov Report

Merging #1774 (60ad0a2) into master (bb689ff) will decrease coverage by 24.01%.
The diff coverage is 70.00%.

@@             Coverage Diff             @@
##           master    #1774       +/-   ##
===========================================
- Coverage   74.32%   50.32%   -24.01%     
===========================================
  Files          81       80        -1     
  Lines        5866     5846       -20     
===========================================
- Hits         4360     2942     -1418     
- Misses       1506     2904     +1398     

Files Changed	Coverage Δ
ext/IncrInfrDiffEqFactorExt.jl	0.00% <0.00%> (ø)
src/services/CalcFactor.jl	54.77% <ø> (-15.93%)	⬇️
src/services/EvalFactor.jl	69.06% <ø> (-8.29%)	⬇️
src/services/FactorGraph.jl	62.44% <40.00%> (-12.01%)	⬇️
src/services/GraphProductOperations.jl	87.09% <83.33%> (-2.56%)	⬇️
src/services/ApproxConv.jl	56.97% <100.00%> (-40.73%)	⬇️
src/services/SolveTree.jl	93.97% <100.00%> (ø)
src/services/SolverUtilities.jl	52.00% <100.00%> (ø)

... and 47 files with indirect coverage changes

📣 We’re building smart automated test selection to slash your CI/CD build times. Learn more