feat: Change molarity concentration to molality concentration

[frankfeifan]

This PR changes convention of the concentration fed into HPCReact from molarity [mol/m3 of solution] or [mol/L of solution] to molality [mol/kg of solvent] using a factor called solventMassPerSolutionVolume (not solventDensity anymore) to be consistent with convention.

This is a preparatory step for introducing the activity model. So now, HPCReact is treated as a strictly molality-based library, and every conversion to a volumetric basis happens on the GEOS side through a single factor.

Specific changes include:

• Switched the concentration passed to HPCReact from molarity to molality, converting back at every caller: accumulation, advective and diffusive flux, Dirichlet and source terms, and their thermal counterparts.

• Introduced solventMassPerSolutionVolume [kg/m3] as an optional input on the reactive fluid model (default 1000) to serve as that conversion factor. For now, it is prescribed and kept constant, but ideally, we want to make it a function of P, T, and chemical composition. Specific implementation of methods and where these updates will be launched (GEOS or HPCReact) needs more discussions and deserves another PR.

• Sourced the solventMassPerSolutionVolume from the solver's single reactive fluid model rather than a per-subregion scan, and turned the previously silent single-fluid-model assumption into an explicit input error, ensuring a single fluid model defined across all regions for a uniform dof layout.

• Restored the finite-difference Jacobian test to dt = 1.0, since the reduction to 0.01 was a workaround for the inflated reaction term.

• Rebaseline

Note: No HPCReact change is required, as the conversion factor lives in GEOS now tentatively.

[jafranc]

Looks great 👍