Improve ESTER chemistry fixing freeeos implementation

[ctmbl]

Roadmap:

• remove old initial_composition funcion and introduce new star2d::init_metal_mix (85d8c3d)
• fix multiple calls (089cad0, 5923d19) to star2d::fill and star2d::init_comp and ensure it (85d8c3d)
• add -mixt <mixt> flag, where <mixt> is either a default mixture (GN93, GS98...) or a file describing a custom one (via relative metal fractions of Z)
  • compatible with -Z
  • verify that the sum of fractions ~= 1, warning if not
• write star2d.comp attribute to hdf5 file (easy)
• read composition (H, He3, He4 and metals) from hdf5 files, providing retro compatibility (harder)
• add -composition <comp.cfg> describing an absolute metal composition in log(n_i/n_H)
  • exclusive with -Z or -mixt
  • Z is computed from the config
  • metals absolute fractions are based on X value through log(n_i/n_H)
• ensure compatibility of -composition, -Z or -mixt with -i <input model>
  • -mixt and -i: keep same Z as input model and change metal mixture
  • -Z and -i: keep same mixture and change metals absolute fractions, Y compensate the surplus/loss of metals to ensure X+Y+Z=1
  • ((-mixt and -Z) OR -composition) and -i: keep nothing, change mixture and Z, Y compensate

TODO:

• fix eos:freeeos which currently is the only one to use star2d.comp-attribute internal metal composition (other uses comp.X|Y|Z() methods) but uses badly (dpeends on basically the whole roadmap) (2f750de, ...)
• check that ester_info star1d star2d (optionally star*dR and ester_evol) works correctly
• use comp attribute in: eos_freeeos, eos_ideal, eos_idealrad, nuc_*
• improve atm_onelayer to avoid calling eos resolution inside the loop (?)
• update python interface, if needed
• Resolve all TODOs in the code
• Store comp in HDF5 file in a more compressed/smart manner (most of metal fields are: uniform or zero), if considered relevant, at least compute the theoretical memory save and write about it in ESTER doc, could be relevant for a huge number of models

Notes:

• Hard-coded supported metal list in star.h
• User is forced to give fractions of every isotop, no hidden hard-coded isotopic fractions except the default one
• Hard-coded He isotopic ratio in star.h
• Mass of every isotop is hardocded (should be changed?)
• step 4 (read HDF5) of the roadmap is needed to make star2d works as it requires to be generated from a 1D model

@rieutord

[ctmbl]

I'll publish here, and edit the TODO list whenever I progress in this project!