Oscillating input power

[trevilo]

This PR collects some recent minor extensions and robustness improvements, including the following:

1. Options to control an oscillating power for the cycle-avg-joule-coupled solver. Specifically the options are

cycle-avg-joule-coupled/oscillating-power = true # boolean to tell solver to use oscillating power, defaults to false
cycle-avg-joule-coupled/input-power-amplitude = 2000.0 # amplitude of oscillation, in Watts
cycle-avg-joule-coupled/power-period = 5000 # period of oscillation, in time steps

2. Minor restart and averaging fixes (e.g., allowing variable order restart with LteThemoChem)
3. Options to clip the temperature field for LteThermoChem and ReactingFlow. Specifically, for reacting,

loMach/reacting/clip-temperature = true # use the clipping, defaults to false
loMach/reacting/min-temperature = 250.0 # set values below 250 to 250 (in Kelvin)
loMach/reacting/max-temperature = 13000.0 # set values above 13000 to 13000 (in Kelvin)

4. An option (use loMach/tomboulides/iorho_gf = true, default behavior is false) to treat the coefficient 1/density as a GridFunctionCoefficient in Tomboulides. This can improve robustness for under-resolved cases. Currently only implemented for 3D.
5. An option to filter the thermal divergence field (use loMach/ltethermo/filter-Q = true, defaults to fales) using a second order differential filter (Qf - \nabla \cdot (\ell^2 Qf) = Q, where Q is the original thermal divergence, Qf is the filtered version, and ell is the mesh resolution length scale. This option is only currently available for LteThermoChem.

Items 4 and 5 should be considered experimental, but are included because of their potential to improve robustness for marginally resolved cases.