various filling tools: output datatype depends on flat_increment/small_num

[jfbourdon]

Contrary to what the tool documentation was saying, the output raster was always Float32 instead of Float64 causing an issue if the flat increment value was determined by the tool, but too small for Float32. Would prevent issues like #106.

Now, the output datatype is set to Float64 if the user doesn't provide any flat increment value. However, if the user provides a value, the output datatype will be the same as the input assuming that user knows what he is doing.

[jfbourdon]

I also made similar changes to BreachDepressions.

[jfbourdon]

Other similar changes in FillDepressions, FillDepressionsWangAndLiu and FillDepressionsPlanchonAndDarboux.
Output DEM in FillBurn is now always promoted to Float64 because the small increment is always calculated and may often be too small for Float32.

Small question, why in FillDepressionsPlanchonAndDarboux the output nodata value is always set to -32768 instead of keeping the original one?

whitebox-tools/whitebox-tools-app/src/tools/hydro_analysis/fill_depressions_planchon_and_darboux.rs

Lines 254 to 258 in c9fa7a6

	let nodata_output = -32768.0f64;
	let large_value = f64::INFINITY;
	let mut output = Raster::initialize_using_file(&output_file, &input);
	output.configs.data_type = DataType::F64;
	output.configs.nodata = nodata_output;