Fix many errors that occur on 32-bit systems under UBSan

[JohnoKing]

This commit fixes fixes a large number of runtime errors that occur when ksh is compiled with -fsanitize=undefined on Linux 32-bit using GCC.

Noted changes:

• Switched out long and double for Sflong_t and Sfdouble_t in many places to fix runtime errors under UBSan. (In certain places (u)int64_t was used to specify exactly 64-bits rather than intmax_t.) The following errors should no longer occur:

src/cmd/ksh93/sh/macro.c:2489:9: runtime error: left shift of 1 by 31 places cannot be represented in type 'long int'
src/cmd/ksh93/sh/args.c:252:4: runtime error: left shift of 1 by 31 places cannot be represented in type 'long int'
src/cmd/ksh93/sh/args.c:272:4: runtime error: left shift of 1 by 31 places cannot be represented in type 'long int'
src/cmd/builtin/pty.c:355:12: runtime error: left shift of 12 by 28 places cannot be represented in type 'long int'
src/cmd/ksh93/sh/args.c:175:60: runtime error: left shift of 1 by 31 places cannot be represented in type 'long int'
src/cmd/ksh93/sh/xec.c:2359:4: runtime error: signed integer overflow: 1734423994 - -1213168416 cannot be represented in type 'long int'
src/cmd/builtin/pty.c:355:12: runtime error: left shift of 12 by 28 places cannot be represented in type 'long int'
src/cmd/ksh93/sh/args.c:607:10: runtime error: left shift of 1 by 31 places cannot be represented in type 'long int'
src/cmd/ksh93/sh/args.c:622:38: runtime error: left shift of 1 by 31 places cannot be represented in type 'long int'

• For mamake, use unsigned long long to avoid requiring <stdint.h>.

• Remove !timeofday code. All modern systems have gettimeofday, which makes this code unused and an unnecessary maintenance burden.

• Define _TIME_BITS=64 to fix integer overflows in the time keyword and likely elsewhere (ksh assumes time_t is 64-bits).

• Moved the AST_LC_* bitflags into the 32-bit range.

• Don't assume long long and intmax_t are the same. While intmax_t is in practice the same as long long, this is not guaranteed by the C standard (it's permitted for these types to differ). For correctness, only use %jd and %ju with Sflong_t (aka intmax_t), and enforce this distinction in SFIO's handling of %ll and %j. (The extra if statements are optimized away on platforms where the types are identical.)

• Use suseconds_t for tv->usec and enforce 64-bit time_t in the libast feature tests.

• Casted more PIDs to Sflong_t when using PIDs with format specifiers.

• Use the %zu format specifier with size_t vars rather than unnecessary casts.

[McDutchie]

Thanks for all this work.

The use of long long in this PR means we now officially abandon C90 and require C99. I am okay with this; see #777.

[McDutchie]

• Remove !timeofday code. All modern systems have gettimeofday, which makes this code unused and an unnecessary maintenance burden.

Looks like gettimeofday was removed in POSIX-1.2024, so we probably should not rely on having it. POSIX-1.2017 marked it as obsolescent and recommends using clock_gettime instead. In our code it would probably be best to use tvgettime from libast.

Since this is unrelated to fixing UBSan errors, I think this would be best dealt with in a separate issue/commit.