fix: control equation in the thermal single-phase well residual norm

[victorapm]

1. The thermal single-phase well residual norm computed a normalizer for the control rows but never added their contribution, so Newton could report convergence while the well rate or BHP target was not met. When a rate-controlled well opened after a shut-in period, a line search could leave it at a fraction of the scheduled rate for the rest of the run, and a restarted run, which rebuilds the well state, followed a different path. Count the control rows as the isothermal kernel does.

Fixes #4036

2. There was an error in the residual scaling for well equations in thermal. This was introduced in the refactor merge. When the well boundary condition is BHP, the wells inactive rate constraint to use for the residual scaling in the component balance equations.

[tjb-ltk]

@victorapm there's another fix PR 4146.
src/coreComponents/physicsSolvers/fluidFlow/wells/kernels/ThermalCompositionalMultiphaseWellKernels.hpp
What do you think about including in this PR ?

[victorapm]

Hi Tom, sounds a good idea to me. Feel free to push to this PR!