Define typemin/typemax for BigInt

[omus]

I think it would be nice to have typemin/typemax defined for all Integer subtypes. I don't think it unreasonable to say that any custom type which is a subtype of Integer should have these methods defined for that type.

This concept was thought of this while working on speeding up some generic code in FixedPointDecimals which used typemax on Integer types. While creating this PR I discovered hastypemax which works but I think it should be eventually be deprecated as the applicable(typemax, T) fallback is not very performant for custom types.

[nalimilan]

Returning nothing is rather weird here. I'd rather make applicable fast... See also #16422.

[StefanKarpinski]

Returning nothing here is definitely weird. I agree with @nalimilan that making applicable faster would be much preferable. Should be possible to have zero overhead in most situations.

[simonbyrne]

I still think it should be AlephNull...

[JeffBezanson]

I agree returning nothing doesn't help much, since there is still a special case you have to check for.

[omus]

If we can make applicable fast than I agree it is preferable to this solution. For now we'll have a special case in FixedPointDecimals until that time.