feat: correct no-loss-of-precision false positives with e notation (#20187)

* fix: correct `no-loss-of-precision` false positives with `e` notation

* add decimal point if missing

* use option to always parse input as a float

* add invalid test case

Author: fasttime