S2164: use BigDecimal if expected type is float

[algomaster99]

As of now, the processor for S2164 is incomplete because of the following case:

         float a = 16777216.0f;
         float b = 1.0f;
-        float c = a + b; // Noncompliant, yields 1.6777216E7 not 1.6777217E7
+        float c = (double) a + (double) b;

See #570 for more details.

However, it can be made fully-fixable if we convert float to java.math.BigDecimal and cast one of the operands to BigDecimal as well.