Provide default thresholds for Float32

[patrickersing]

PR #38 introduces two new thresholds threshold_desingularization and threshold_partially_wet.
Default values for these thresholds are currently provided by a fixed value. As we do not have experience with other number formats, the default values are only available for Float64 (see #38 (comment)).

For other number formats thresholds can be set manually through the equation struct, e.g.
equations = ShallowWaterMultiLayerEquations1D(gravity_constant=9.812f0, rhos=1f0, threshold_limiter=5eps(), threshold_desingularization=1f-6, threshold_partially_wet=1f-4)

However, we should also test and introduce good default values for Float32. The default value threshold_partially_wet=1e-4 can probably also be used in single precision, but threshold_desingularization=1e-10 needs to be adjusted.