Floating-point number support

[fmease]

Support for IEEE-754 floating point support and also a more type-safe interface without NaN`and/or infinity.

• [ ] extend the lexer to tokenize e.g. 1.2, 0.000_000, 1e4, 1e-3_1 as numbers
• [ ] the foreign types core.float32.Float32 and core.float64.Float64 with neither NaN nor infinity (unsure about the latter)
• [ ] the foreign types core.float32.raw.Float32 and core.float64.raw.Float64 with NaN and infinity (unsure about the naming: Raw-FloatN?)
• [ ] arithmetic, ordering and conversion operations for all of those
• [ ] common constants for all of these (pi, e, tau, …)
• [ ] number literals for float (scientific notation) with the "literal marker attributes" @Float32, @Float64, @Raw-Float32, @Raw-Float64