feat(sim): add variants/sim/ — real companion_radio firmware on native + Emscripten

New board variant compiling the unmodified MyMesh/UITask/DataStore app
logic against real mesh::Radio/MainBoard/RTCClock/RNG interfaces, for
running the actual firmware outside embedded hardware:

- Native (plain g++, platform = native): ASCII-art display over stdout,
  stdin-driven input, local-disk-backed DataStore/IdentityStore.
- Emscripten/WASM (variants/sim/build_wasm.sh, since PlatformIO's native
  platform force-overrides any CC/CXX toolchain override back to system
  clang++): canvas-backed display, IDBFS-backed persistence across page
  reloads, JS-callable input via sim_enqueue_key(), emscripten_set_main_loop.

Real rweather/Crypto (AES128/SHA256/Ed25519) vendored unmodified and
proven working on both targets. variants/sim/web/index.html is a bare
verification harness, not the polished website embed.

variants/sim/sim_main.cpp

@@ -0,0 +1,112 @@
+// Process/runtime entry point for the sim build. Real Arduino cores provide
+// their own main() (call setup() once, then loop() forever); platform=native
+// has no such core, so this is that main() for both Phase 1's native target
+// and Phase 2's Emscripten target -- the two are different enough (a native
+// process owns its own loop and a real stdin tty; a wasm module in a browser
+// tab must hand control back to the browser's event loop between ticks, and
+// has no stdin at all) that they get fully separate #ifdef __EMSCRIPTEN__
+// branches below rather than one branch trying to cover both.
+#include <Arduino.h>
+#include <cstdio>
+#include <cstdlib>
+
+// Defined by examples/companion_radio/main.cpp (Arduino sketch convention:
+// no header declares these, every board's own entry point just forward-
+// declares and calls them).
+void setup();
+void loop();
+
+#ifdef __EMSCRIPTEN__
+// ---------------------------------------------------------------------
+// Phase 2: Emscripten/browser entry point.
+#include <emscripten.h>
+#include "SimFS.h"   // sim_fs_mount_idbfs()
+
+// One tick of the real app's cooperative loop -- identical body to
+// examples/companion_radio/main.cpp's own loop() (the_mesh.loop();
+// sensors.loop(); ui_task.loop(); rtc_clock.tick(); ...), so this wrapper
+// adds nothing of its own; it exists only because emscripten_set_main_loop
+// wants a void(void) function pointer, and passing `loop` directly would
+// also work here, but naming it makes the call below self-documenting.
+static void sim_main_loop_tick() {
+  loop();
+}
+
+// Called back from JS (see SimFS.h's sim_fs_mount_idbfs()) once the async
+// IndexedDB -> MEMFS pull finishes. This is where the app's boot sequence
+// actually starts -- deliberately NOT in main() itself, since main() must
+// return (or hand off via emscripten_exit_with_live_runtime()) long before
+// this callback ever fires.
+extern "C" EMSCRIPTEN_KEEPALIVE void sim_idbfs_ready() {
+  setup();
+  // fps=0, simulate_infinite_loop=1: let the browser's own
+  // requestAnimationFrame cadence drive ticks (Emscripten's documented
+  // recommendation for anything drawing to a <canvas>) rather than a fixed
+  // interval -- ties the sim's tick rate to the actual display refresh, and
+  // it's throttled/paused for free by the browser when the tab is hidden or
+  // backgrounded, which a manual setInterval(..., fixed_ms) wouldn't get.
+  emscripten_set_main_loop(sim_main_loop_tick, 0, 1);
+}
+
+int main() {
+  printf("MeshCore sim (wasm) starting -- mounting IDBFS at /sim_data...\n");
+  // "/sim_data" must match the relative "./sim_data" DataStore store(sim_fs,
+  // ...) in examples/companion_radio/main.cpp resolves to -- see the long
+  // comment on sim_fs_mount_idbfs() in SimFS.h for why those are the same
+  // filesystem node under Emscripten's default cwd ("/").
+  sim_fs_mount_idbfs("/sim_data");
+  // Keep the runtime alive after main() returns instead of tearing it down
+  // (the default for a `main()` that returns under Emscripten) -- the real
+  // boot sequence hasn't happened yet, it's waiting on sim_idbfs_ready()
+  // above, an async JS callback that fires well after main() itself is done.
+  emscripten_exit_with_live_runtime();
+  return 0;
+}
+
+#else
+// ---------------------------------------------------------------------
+// Phase 1: native (host process) entry point -- unchanged from the
+// original Phase 1 implementation. Puts the terminal into raw/
+// non-canonical mode so UITask.cpp's SIM_PLATFORM stdin-poll branch (see
+// UITask::loop()) receives individual keystrokes immediately instead of
+// once per Enter-terminated line.
+#include <unistd.h>
+#include <termios.h>
+
+static struct termios g_orig_termios;
+static bool g_termios_saved = false;
+
+static void restoreTerminal() {
+  if (g_termios_saved) tcsetattr(STDIN_FILENO, TCSANOW, &g_orig_termios);
+}
+
+static void setupTerminal() {
+  if (!isatty(STDIN_FILENO)) return;   // piped/redirected stdin -- leave alone
+  if (tcgetattr(STDIN_FILENO, &g_orig_termios) != 0) return;
+  g_termios_saved = true;
+  atexit(restoreTerminal);
+
+  struct termios raw = g_orig_termios;
+  raw.c_lflag &= ~(ICANON | ECHO);   // no line buffering, no local echo
+  raw.c_cc[VMIN] = 0;
+  raw.c_cc[VTIME] = 0;
+  tcsetattr(STDIN_FILENO, TCSANOW, &raw);
+}
+
+int main() {
+  setupTerminal();
+  printf("MeshCore sim (native) -- Ctrl-C to quit.\n");
+  printf("Keys: arrows/WASD move, Enter/Space select, Esc/Backspace cancel, n/p next/prev.\n\n");
+
+  setup();
+  for (;;) {
+    loop();
+    // Real boards spin their loop() as fast as the hardware allows too;
+    // this just keeps a native process from pegging a CPU core at 100%
+    // for no benefit -- 1ms is well under any UI timing this app cares
+    // about (refresh/animation intervals are tens-to-hundreds of ms).
+    usleep(1000);
+  }
+  return 0;
+}
+#endif // __EMSCRIPTEN__