Trac #33201: fix Cython warnings for dense float/double matrices.

In our template for dense float/double matrices mod-n, we define a few
variables "zero", "one", and "d" that are passed by reference to C
functions to be overwritten. Cython however is not smart enough to
know that these uninitialized variables are about to be initialized in
C, and throws warnings like

  warning: sage/matrix/matrix_modn_dense_template.pxi:283:19: local
  variable 'd' referenced before assignment

To avoid the warnings, we initialize all of these variables (whose
type is either float or double, depending on the context) to zero.

Author: orlitzky

src/sage/matrix/matrix_modn_dense_template.pxi

@@ -274,7 +274,7 @@ cdef inline celement linbox_det(celement modulus, celement* entries, Py_ssize_t
     cdef ModField *F = new ModField(<long>modulus)
     cdef celement *cpy = linbox_copy(modulus, entries, n, n)
 
-    cdef celement d
+    cdef celement d = 0
     cdef size_t nbthreads
     nbthreads = Parallelism().get('linbox')
 
@@ -293,7 +293,7 @@ cdef inline celement linbox_matrix_matrix_multiply(celement modulus, celement* a
     C = A*B
     """
     cdef ModField *F = new ModField(<long>modulus)
-    cdef ModField.Element one, zero
+    cdef ModField.Element one = 0, zero = 0
     F[0].init(one, <int>1)
     F[0].init(zero, <int>0)
 
@@ -319,7 +319,7 @@ cdef inline int linbox_matrix_vector_multiply(celement modulus, celement* C, cel
     C = A*v
     """
     cdef ModField *F = new ModField(<long>modulus)
-    cdef ModField.Element one, zero
+    cdef ModField.Element one = 0, zero = 0
     F.init(one, <int>1)
     F.init(zero, <int>0)