improve how calc compiles on big endian machines

Fixed bug that caused calc to fail to compile filepos2z() in file.c
on little endian machines for the Debian apcalc package.  Thanks to
Martin Buck (m at rtin-buck dor de) for for fix.

Removed unused macros from zmath.h:

    SWAP_B32_IN_HASH(dest, src)
    SWAP_B16_IN_HASH(dest, src)
    SWAP_B8_IN_HASH(dest, src)
    SWAP_B32_IN_FLAG(dest, src)
    SWAP_B16_IN_FLAG(dest, src)
    SWAP_B8_IN_FLAG(dest, src)

When SWAP_HALF_IN_B32(dest, src), SWAP_B32_IN_FULL(dest, src),
SWAP_B16_IN_HALF(dest, src), SWAP_B32_IN_bool(dest, src),
or SWAP_B32_IN_LEN(dest, src), SWAP_HALF_IN_FILEPOS(dest, src)
is an assignment such as:

    (*(dest) = *(src))

We now case the dest and src pointers to the proper type before
referencing and performing the assignment.

file.c

@@ -1387,7 +1387,7 @@ filepos2z(FILEPOS pos)
 	ret.len = FILEPOS_BITS/BASEB;
 	ret.v = alloc(ret.len);
 	zclearval(ret);
-	SWAP_HALF_IN_FILEPOS(ret.v, &pos);
+	SWAP_HALF_IN_FILEPOS(ret.v, (HALF *)&pos);
 	ret.sign = 0;
 	ztrim(&ret);