Check for signals every 32nd iteration and once per Karatsuba step

With the 64-row batch, Karatsuba base cases (36-70 rows) mostly never
reached the check, so e.g. a 200k-digit product ran for seconds without
handling a signal. A 32-row batch makes every substantial base case
check, and k_mul() additionally checks once per recursion step.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>

Author: eendebakpt

Objects/longobject.c

@@ -113,10 +113,10 @@ maybe_small_long(PyLongObject *v)
  */
 #define HUGE_EXP_CUTOFF 60
 
-/* Check for signals on every 64th iteration `i` of a loop. */
+/* Check for signals on every 32nd iteration `i` of a loop. */
 #define SIGCHECK(i, PyTryBlock)                                     \
     do {                                                            \
-        if ((((i) & 63) == 63) && PyErr_CheckSignals()) PyTryBlock  \
+        if ((((i) & 31) == 31) && PyErr_CheckSignals()) PyTryBlock  \
     } while(0)
 
 /* Normalize (remove leading zeros from) an int object.
@@ -4111,6 +4111,13 @@ k_mul(PyLongObject *a, PyLongObject *b)
             return x_mul(a, b);
     }
 
+    /* x_mul() only checks for signals every 32nd row and the base cases of
+       the recursion below can have fewer rows than that, so check here,
+       once per Karatsuba step. */
+    if (PyErr_CheckSignals()) {
+        return NULL;
+    }
+
     /* If a is small compared to b, splitting on b gives a degenerate
      * case with ah==0, and Karatsuba may be (even much) less efficient
      * than "grade school" then.  However, we can still win, by viewing