fix(HIGH-01): add token-count guard and embed_lookup bounds check

- train_large.m: early exit if n_tokens < SEQ+1 prevents arithmetic
  underflow and downstream index corruption
- stories_cpu_ops.h: clamp tok >= VOCAB to 0 in embed_lookup() to
  prevent heap-buffer-overflow

ref: docs/reports/security-audit-2026-03-02.md HIGH-01

training/stories_cpu_ops.h

@@ -112,6 +112,7 @@ static float cross_entropy_loss(float *dlogits, const float *logits, const uint1
 static void embed_lookup(float *x, const float *embed, const uint16_t *tokens, int dim, int seq) {
     for (int t = 0; t < seq; t++) {
         int tok = tokens[t];
+        if (tok >= VOCAB) { tok = 0; }  // HIGH-01: clamp invalid token -> position 0
         for (int d = 0; d < dim; d++) {
             x[d*seq + t] = embed[tok*dim + d];
         }

training/train_large.m

@@ -296,6 +296,10 @@ int main(int argc, char *argv[]) {
         uint16_t *token_data = (uint16_t*)mmap(NULL, data_len, PROT_READ, MAP_PRIVATE, data_fd, 0);
         if (token_data == MAP_FAILED) { printf("mmap failed\n"); return 1; }
         size_t n_tokens = data_len / 2;
+        if (n_tokens < (size_t)SEQ + 1) {
+            fprintf(stderr, "Token file too small: %zu tokens, need >%d\n", n_tokens, SEQ + 1);
+            return 1;
+        }
         printf("Token data: %zu tokens (%.1f MB)\n", n_tokens, data_len/1e6);
 
         // Gradient buffers shared across layers (reused each step)