Trac 12866: improve asymptotic of monodromy_weights()

This changes monodromy_weights() so that it doesn't need to compute
all O(n^2) ideal products -- rather only O(n), with the corresponding
asymptotic improvement.

This is important for large dimensions where one doesn't use the brandt
matrix method to compute hecke operators, but rather the 'direct' method
which doesn't require to compute all the O(n^2) products either.

src/sage/modular/quatalg/brandt.py

@@ -1413,8 +1413,9 @@ def monodromy_weights(self):
         #     c * theta(R) = a[0] + a[1]*q + ...
         #
         # we recover e = a[1]/a[0] regardless of normalization.
-        bv = self._brandt_series_vectors(2)
-        w = tuple(bv[i][i][1]/bv[i][i][0]/2 for i in range(len(bv)))
+        orders = self._ideal_products(diagonal_only=True)
+        thetas = (R.theta_series_vector(2) for R in orders)
+        w = tuple(a[1]/a[0]/2 for a in thetas)
         self.__monodromy_weights = w
         return w