Supporting smaller floats (i.e 4 to 24 bits) with ArbNumberics.jl package

[tkgunaratne]

Floating point numbers represented with smaller number of bits (i.e. Half-precision (float16 - 10 bit mantissa), Brain-Float16 (7-bit mantissa) are being supported by some hardware vendors (e.g. Intel Agilex FPGAs). Is there a plan to accommodate these formats and other general formats in ArbNumerics.jl package?