Remove some of the workarounds for the tests by inputing the right numbers
(without impliciit conversion from int to long double).

The previous commit message talked about reading numbers, but the i

Remove some of the workarounds for the tests by inputing the right numbers
(without impliciit conversion from int to long double).

The previous commit message talked about reading numbers, but the issue
(loss of precision due to too little digits) actually occurred when printing
the input to be read by bc.

show more ...

arm64 has real IEEE quads, it needs more digits when reading the
bc results otherwise truncation and loss of precision occurs, making the
test fail.

Test failure on arm64 noted by anton@ and bluhmn@

fix decl of main

comment in Makefile, small tweak in -l test

better bounds

rough test of the bc(1) math lib