Replace the float-point-type parameter used internally to select float, double
or long double in computation (a template parameter for all solver) with a
simple preprocessor parameter BARYONYX_FLOATINGPOINT_TYPE. Must be equal to 0, 1
or 2 for respectively (float, double, long double).

Due to `Windows.h`, we need to extra paranthesis the min/max function
access.

Previously, only the exhaustive solver was used and could lead to NOEM.
This patch allows to switches from B&B solver, exhaustive solver and a
local greedy solver according to the complexity of the constraint.

(Reference: #35).

(Reference: #35).

(Reference: #35).

Add missing parameters

Use codecov to procudes coverage reports

This reverts commit eeea6927.