Provide class FloatValue for storing arbitrary-size floating-point values

This MR takes the existing work on having a class for handling arbitrary-precision floating-point values and their arithmetic in CPAchecker and finishes it. Now the implementation is purely in Java and no longer relies on a C library, which makes it available on all platforms. The new class is also used by the float-literal expressions.

More detailed description of the MR is in !136 (comment 1891646858). The main issue this MR solves is #349. It should also solve #1072. Several other issues related to float and component: util.floatingpoint may now also be resolved, for example maybe #500. #1074 and #575 are certainly fixed, and the tests introduced or #575 are now enabled. Some cleanup that does not affect consumers of this code needs to be done afterwards, e.g. #1315.