Use fallback implementation for min, max

[Socob]

min and max are currently explicitly defined for AbstractQuantity arguments and have their own separate re-implementation of uconvert for some reason. However, it turns out that this is unnecessary since the standard fallback implementations of min and max already work perfectly fine for AbstractQuantity, because comparison is also defined. Thus, this simply removes these extra method definitions, which, as a bonus, also provides all of the error checking added in #648 to min and max as well.

[sostock]

Great! This also fixes another bug, because the deleted methods used < instead of isless:

Before (wrong):

julia> min(NaN*u"m", 1.0u"m")
1.0 m

julia> min(1.0u"m", NaN*u"m")
NaN m

After (correct):

julia> min(NaN*u"m", 1.0u"m")
1.0 m

julia> min(1.0u"m", NaN*u"m")
1.0 m

It would be good to add tests for that as well.

[Socob]

@sostock Ah yes, I didn’t even notice that. Tests added!