Fix definitions of type_min and type_zero in xr_types.h

[AMS21]

While looking at the code inside xr_types.h I've notices some very strange definitions which are used later in the code to define constants.

// Type limits
template <typename T>
constexpr auto type_max = std::numeric_limits<T>::max();

template <typename T>
constexpr auto type_min = -std::numeric_limits<T>::max();

template <typename T>
constexpr auto type_zero = std::numeric_limits<T>::min();

type_min is not the minimum value of integer types it is actually the minimum + 1. This only works for floating point types.
type_zero is the actual minimum of a type. But is only 0 for unsigned types. For signed types its a negative value.

Thus we have some very strange constants.

constexpr int int_min = type_min<int>;         // actually -2147483647 not -2147483648
constexpr int int_zero = type_zero<int>;       // actually -2147483648 not 0
...
constexpr float flt_zero = type_zero<float>;   // actually 1.175494e-38 not 0.0
...
constexpr double dbl_zero = type_zero<double>; // actually 2.225074e-308 not 0.0

Fixes #195

[Xottab-DUTY]

This is related to #195.

One of the reasons why this inconsistency still wasn't fixed is that I really don't know if everything will work correct after the change.

Knowing the codebase, some code logic could depend on this inconsistency. The situation when some code was broken because another unrelated code has been changed happened many times

[AMS21]

Okay I've now added type_lowest and accompanying constants and changed all usages of these constants across the code base to keep the old behavior.

Making these changes I'm a bit suspicious about some of the uses of flt_zero now flt_min. Some of them look to me like they actually meant to check for 0.

[AMS21]

Can this ever be true?

[AMS21]

Similar here this would always be true since flt_min is negative