Build error on Big Endian

Hi,

While building tfx-bsl which needs ZetaSQL on big endian, observed below error:
```
external/com_google_zetasql/zetasql/common/multiprecision_int_impl.h:317:33: note: candidate: template std::array::Uint, n1> zetasql::multiprecision_int_impl::Convert(const std::array::Uint, n2>&, bool)
inline std::array, n1> Convert(const std::array, n2>& src,
^~~~~~~
external/com_google_zetasql/zetasql/common/multiprecision_int_impl.h:317:33: note: template argument deduction/substitution failed:
external/com_google_zetasql/zetasql/common/multiprecision_int_impl.h:565:49: note: candidate expects 2 arguments, 1 provided
Array32 dividend32 = Convert<32, n * 2, 64, n>(dividend);
~~~~~~~~~~~~~~~~~~~~~~~~~^~~~~~~~~~
external/com_google_zetasql/zetasql/common/multiprecision_int_impl.h:570:42: error: no matching function for call to 'Convert<64, 2, 32, (2 * 2)>(Array32&)'
*quotient = Convert<64, n, 32, n * 2>(quotient32);
~~~~~~~~~~~~~~~~~~~~~~~~~^~~~~~~~~~~~
```

This is resolved with below patch:
```
diff --git a/zetasql/common/multiprecision_int_impl.h b/zetasql/common/multiprecision_int_impl.h
index 55ce01a..52d7c02 100644
--- a/zetasql/common/multiprecision_int_impl.h
+++ b/zetasql/common/multiprecision_int_impl.h
@@ -315,7 +315,7 @@ inline void Copy(const Uint* src, int src_size, Uint* dest,
// allow_optimization is used only for testing.
template
inline std::array, n1> Convert(const std::array, n2>& src,
- bool negative) {
+ bool negative = false) {
std::array, n1> res;
Uint extension = negative ? ~Uint{0} : 0;
#ifndef ABSL_IS_BIG_ENDIAN
```

As contributions are not allowed via PR, could someone help in reviewing and fixing the issue in source code?

Start in zetasql/common/multiprecision_int_impl.h at the Convert template and its calls around lines 565 and 570. Reproduce the tfx-bsl build on a big-endian target, then verify that the reported template deduction errors are gone and the multiprecision code still builds successfully.

Written by the indexing model from the issue text.