colorspace: don't nuke HDR10 minimum in pl_color_space_infer
We only wanted to sanitize values here, not strip them for PQ (as pl_color_space_nominal_peak_ex implies). Probably the smarter solution would be to just add explicit checks and only infer missing values, but since in practice only the peak value is ever that horribly broken, just skip overriding the minimum value if it's present.