Floating-point support to SSA interpreter and bugfixes

[bmwangmh]

Main contributions:

• SSA interpreter now supports floating-point numbers. See da1ed53.
• A preliminary purity anaylsis has been done, now dead calls to pure functions are also eliminated. See ca7c072 and 7e127c4.

Test cases:

• Basic floating-point instructions: floatpoint01, floatpoint02.

Bugfixes:

• The compiler now correctly handles closure self-references, allowing a closure to correctly call itself. Test closure02 has passed. A new test closure05 was added as a simplified version of closure02. See 7e6cb22.
• An omitted Assign rd _ was added when dealing with Psetbytesitem See db0df65.
• A redundant = in the output of AssignFP was removed. The rule for deciding if an operation is unsigned in the interpreter has been changed to be based on rs instead of rd. See 957528a.
• Logical Not now correctly uses logical not instead of bitwise ~ in the interpreter. See 6f566d8.