liquidTest not cleaning up database -- causes problems for downstream tests

[sevyharris]

Bug Description

I've got a PR changing how we parse comments, and some of the tests are showing that there's still solvation data in the database that isn't getting cleared. https://github.com/ReactionMechanismGenerator/RMG-Py/actions/runs/26604721808/job/78396963318?pr=2963

My test estimates thermo for the adsorbate CH3X, and the resulting thermo comment contains:
Gas phase thermo for [CH3] from Thermo library corrected for liquid phase: primaryThermoLibrary + radical(CH3)

How To Reproduce

run pytest test/rmgpy/solver/liquidTest.py minTest.py using this example minimal test minTest.py.txt which passes by itself and fails when paired with the liquidTest

Expected Behavior

The thermo comment shouldn't have any references to liquid phase.

I didn't realize that pytest doesn't clear memory between test files. I'd been writing tests with the assumption that they're starting from a blank slate, but this goes to show that tests should clean up the database after using them, and perhaps use some more defensive clearing at the beginning since we have so many contributors and that's a hard thing to enforce.

[sevyharris]

Script for brute-force testing what's breaking uncertaintyTest. Can run this later when I have more time.

#!/bin/bash

for test_file in $(find test/rmgpy/ -name "*.py" -not -path "*/uncertaintyTest.py"); do
    echo "Testing: $test_file"
    result=$(pytest "$test_file" test/rmgpy/tools/uncertaintyTest.py -x -q --tb=no 2>&1 | tail -1)
    echo "$result"
    if echo "$result" | grep -q "failed"; then
        echo "^^^ CULPRIT FOUND: $test_file"
    fi
done

[sevyharris]

You can fix it by adding rmgpy.rmg.main.solvent = None to the teardown class.