Add a headless, fixed-step simulation mode for batch experiments

[BTawaifi]

Split out of #7874, now on top of #7834. Works best with the afterburner cooldown fix (#7895) merged first, since afterburner cooldowns otherwise run on the real clock while headless runs simulate many seconds per real second.

An opt-in mode for running single-player missions unattended and fast, for balancing, regression and AI experiments. Nothing changes unless -headless is passed.

Option	Effect
-headless	No window, renderer, audio, speech, joystick, HUD/radar or movies; the stub renderer is used, model and collision data still load. Frames run as fast as the CPU allows on a mission clock that advances only by simulated time (wall-clock APIs are unchanged). Implies -noninteractive and -nomovies.
-fixed-dt <s>	Authoritative step, default 1/30, range [1/65536, 0.25].
-max-sim-time <s>	End the run after this much simulated time (at most 32767, the fixed-point limit).
-sim-result <path>	Write a JSON result at shutdown: mission, seed, simulated and wall seconds, why the run ended (time_limit, player_died, start_failed, mission_exit) and the state of each mission goal.

The fast clock is not a separate option: on its own it would freeze menu timers and break multiplayer timing.

What runs: the normal authoritative frame (game_frame(), game_simulation_frame()), so objects, AI, weapons, collisions, goals and scripting all run. On Frame callbacks normally fire from gr_flip(); they are split out as gr_process_frame_callbacks() and run once per simulated frame. Keyboard and joystick input is not read, but read_player_controls() still runs, so scripted (Lua) flight control, player physics and the player warp-out sequence work.

Unattended flow: with -start_mission, the fiction and command briefing are skipped, a red-alert mission goes through the red alert screen as its Continue button does, the default loadout is committed through commit_pressed() (as for no-briefing missions), and the debriefing is left by quitting. A failed commit ends the run (start_failed). The death popup cannot be answered, so the player's death ends the run (player_died). Other modal popups are logged and answered with the abort answer, unless a script handles On Dialog Frame.

Second commit: under -noninteractive, Lua errors are now printed to stderr before the exception ends the process, as Error() already does. Release builds write no log, so before this a Lua error (e.g. a syntax error in a script table) ended an unattended run with no indication why.

Without -headless the additions are single boolean checks on the normal path (timer read, frame timing, radar_plot_object, popup(), the player-controls block).

Testing (retail sm1-01, Release unless noted):

• time limit, end-mission from a script, and the player killed at 10 s end as time_limit, mission_exit and player_died; without -max-sim-time a player death now ends the run in 1 s (before: never)
• Lua steering under -headless turns the player ship (heading dot 0.39 after 10 s, vs 1.00 without steering and 1.00 on Add an opt-in headless fixed-step simulation mode, seeded RNG, On Ship Damage hook and batch runner #7874's build)
• equal seeds give identical end states (with Make seeded runs reproducible across the RNG families the simulation uses #7893)
• a Lua syntax error now reports 'end' expected ... near '<eof>' on stderr
• the Debug build, with assertions, passes the same runs; the combined branches build with FSO_FATAL_WARNINGS=ON on Windows (MSVC) and Linux (gcc-13, clang-16), and the unit tests pass

Known limits: single player only; without -start_mission (or a script driving the menus) the engine waits at the main hall; long runs with many AI ships can still diverge for equal seeds.