GERG rewrite

[longemen3000]

• GERG now uses the param machinery. it could open the door to parse parameters from files, like TREND does. with that rework, came a lot of other secundary things:
  New functions:
  pack_vectors(SingleParam{Vector{T}} -> PackedVectorSingleParam{T}, stores the vector of vectors in a flat buffer, speeding up access.
  pack_vectors(PairParam{Vector{T}} -> PackedSparsePairParam{T}, vector of vectors in a flat buffer, and in a sparse manner.
  Improved functions:
  -saturation_pressure and crit_pure are faster (we now use a MVector to perform the non linear solve in those cases. maybe going full static could speed up that case further)
  -volume(model,p,T,z,phase=phase) does not allocate (if the eos does not allocate) and is inferrable.
  -SingleParam and PairParam can now use a different storage type than Vector{T}
  new EoS:
• EOS_LNG: source
• LJRef: source (it needs a database of LJ data, supports multiple components via vdw1 mixing)
  there is also commented code of UFTheory. basically LJ and mie monomer data is needed.
  improved and/or fixed EoS:
  -error in IAPWS95 (that went unnoticed for a long time)
  -PropaneRef and IAPWS95 are faster
  -GERG2008 is faster (we need benchmarks), EOS_LNG uses the same code than GERG2008.
  -cubic translations and alphas were optimized. (now activity models are the main bottleneck)
  other
  The project was reestructured to make all the parts as independent as possible. with that, a clear view of what are the internal dependencies is possible:

base --> params -|-> database ----------------|
                 |-> split_model --> methods -|-> models
                 |-> macros ------------------|

The core functionality is in base --> params--> split_model --> methods it is possible to define new EoS with only that chain of files.

we now use a custom AD machinery, the functions ∂2f and ∂2a are faster and in general, the base differentials should make more legible errors, based on Improved ForwardDiff.jl Stacktraces With Package Tags

[codecov-commenter]

Codecov Report

Merging #61 (38e80a0) into master (b0be126) will decrease coverage by 0.20%.
The diff coverage is 87.57%.

@@            Coverage Diff             @@
##           master      #61      +/-   ##
==========================================
- Coverage   88.60%   88.39%   -0.21%     
==========================================
  Files         123      127       +4     
  Lines        7168     7396     +228     
==========================================
+ Hits         6351     6538     +187     
- Misses        817      858      +41     

Impacted Files	Coverage Δ
src/Clapeyron.jl	100.00% <ø> (ø)
...rs/multicomponent/tp_flash/RachfordRicetp_flash.jl	88.09% <ø> (ø)
src/models/Activity/COSMOSAC/COSMOSAC02.jl	100.00% <ø> (ø)
src/models/Activity/COSMOSAC/COSMOSAC10.jl	100.00% <ø> (ø)
src/models/Activity/COSMOSAC/COSMOSACdsp.jl	86.84% <ø> (ø)
src/models/Activity/COSMOSAC/utils.jl	100.00% <ø> (ø)
src/models/Activity/NRTL/NRTL.jl	100.00% <ø> (ø)
src/models/Activity/UNIFAC/UNIFAC.jl	100.00% <ø> (ø)
src/models/Activity/UNIFAC/variants/PSRK.jl	100.00% <ø> (ø)
src/models/Activity/UNIFAC/variants/VTPR.jl	88.23% <ø> (ø)
... and 110 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update b0be126...38e80a0. Read the comment docs.