cgo: add support for C.float and C.double

They are not necessary in TinyGo because they always map to float32 and
float64, but it's a good idea to add them regardless for compatibility
with existing software.

(Now I think about it, perhaps it would have been better to require
explicit casts here just in case we want to support some really weird C
system, but then again we even force 64-bit double on AVR even though
avr-gcc defaults to 32-bit double).

testdata/cgo/out.txt

@@ -32,7 +32,8 @@ struct: true 256 -123456 +3.140000e+000
 array: 5 6 7
 union: true
 union s: 22
-union f: +3.140000e+000
+union f (C.float): +6.280000e+000
+union f (float32): +3.140000e+000
 union global data: 5 8 1
 union local data:  5 8 1
 union s: true