Port ThermoML parsing, add data entry types

[mattwthompson]

Closes #2
Closes #25
Closes #26

• Require RDKit
• Drop ThermoML plugin machinery (in favor of built-in solutions)
• Drop most Evaluator machinery in favor of simpler solutions
  • No ThermodynamicState, just attributes for temperature and pressure
  • Use None in place of UNDEFINED
  • Remove Attribute / AttributeClass
  • Port some pre-Pydantic classes into Pydantic
• Define Arrow schema for entry types
• Look at how substance submodule is used (Re-use Evaluator's Substance classes? #40) but not necessarily refactor it

[mattwthompson]

It turns out the implementation in Evaluator requires each property to be fully set up in openff/evaluator/properties/ before the corresponding ThermoML data can be parsed

[mattwthompson]

This test is failing (and probably the cause of other failures):

FAILED dimsim/_tests/datasets/test_thermoml.py::test_load_property_types[single_dhvap.xml-expected2] - Exception: No properties parsed

Dropping into a debugger indicates it's deeper:

dimsim/_tests/datasets/test_thermoml.py::test_load_property_types[single_dhvap.xml-expected2] > /Users/mattthompson/software/dimsim/dimsim/datasets/thermoml/thermoml.py(2041)from_xml()
   2040             import ipdb; ipdb.set_trace()
-> 2041             properties = _PureOrMixtureData.from_xml_node(node, namespace, compounds)
   2042

ipdb> node
<Element '{http://www.iupac.org/namespaces/ThermoML}PureOrMixtureData' at 0x16a186de0>
ipdb> namespace
{'ThermoML': 'http://www.iupac.org/namespaces/ThermoML'}
ipdb> compounds
{1: Compound(smiles=Fc1ccccc1Cl, index=1)}
ipdb> _PureOrMixtureData.from_xml_node(node, namespace, compounds)
ipdb> _PureOrMixtureData.from_xml_node(node, namespace, compounds) is None
True

This is the code most blatantly copied from Evaluator, a.k.a. I understand it the least

Note that similar tests for other properties are passing

[codecov-commenter]

Codecov Report

❌ Patch coverage is 52.31362% with 742 lines in your changes missing coverage. Please review.
✅ Project coverage is 57.83%. Comparing base (c91f015) to head (c417b33).

Files with missing lines	Patch %	Lines
dimsim/datasets/thermoml/thermoml.py	66.58%	273 Missing ⚠️
dimsim/datasets/datasets.py	31.87%	171 Missing ⚠️
dimsim/utils/serialization.py	20.61%	154 Missing ⚠️
dimsim/substances/substances.py	28.65%	122 Missing ⚠️
dimsim/substances/amounts.py	55.55%	20 Missing ⚠️
dimsim/substances/components.py	93.10%	2 Missing ⚠️

❗ There is a different number of reports uploaded between BASE (c91f015) and HEAD (c417b33). Click for more details.

HEAD has 16 uploads less than BASE

Flag	BASE (c91f015)	HEAD (c417b33)
	24	8

Additional details and impacted files

@@             Coverage Diff             @@
##             main      #24       +/-   ##
===========================================
- Coverage   89.66%   57.83%   -31.84%     
===========================================
  Files           4       12        +8     
  Lines         271     1819     +1548     
===========================================
+ Hits          243     1052      +809     
- Misses         28      767      +739     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[mattwthompson]

Decision: let RDKit (but not AmberTools) be a required dependency of dimsim. So no need to test with/without RDKit

Need to:

• Define schema for each data entry type (see similar code in Descent)
• Think more about if substances submodule needs to be ported here (Re-use Evaluator's Substance classes? #40)
• Try out using mapped SMILES when going from ThermoML results to data entry types

Nice to haves:

• look through ported Evaluator code and remove dead/unused code
• Look through some Enums/fancy base classes and see what can be removed for simplicity
  • Example: PropertyPhase enum could be simpler
  • Stretch goal: have coverage 90%+ without adding more tests
• dimsim/utils/serialization.py could probably be replaced by existing Pydantic-based serialization code
• intermediate ThermodynamicState class could be removed

[mattwthompson]

Would be better to switch into different class constructors here, based on the value of measured_property.type_string

For now, keep the tag attribute - even though it's redundant now, it might be useful later on (and it can be dropped later if it is not)