32-bit safety in big random number

[barak]

This patch fixes a compile-time warning and test case failure on
32-bit architectures, including i386.

cd /<<PKGBUILDDIR>>/obj-i686-linux-gnu/tests && /usr/bin/c++   -I/<<PKGBUILDDIR>>/include  -g -O2 -fdebug-prefix-map=/<<PKGBUILDDIR>>=. -fstack-protector-strong -Wformat -Werror=format-security -Wdate-time -D_FORTIFY_SOURCE=2 -fopenmp -Wall -Wpedantic -Wunused-parameter   -std=gnu++11 -o CMakeFiles/ensmallen_tests.dir/cmaes_test.cpp.o -c /<<PKGBUILDDIR>>/tests/cmaes_test.cpp
/<<PKGBUILDDIR>>/tests/callbacks_test.cpp: In function ‘void ____C_A_T_C_H____T_E_S_T____46()’:
/<<PKGBUILDDIR>>/tests/callbacks_test.cpp:397:28: warning: unsigned conversion from ‘long long int’ to ‘size_t’ {aka ‘unsigned int’} changes value from ‘10000000000’ to ‘1410065408’ [-Woverflow]
  397 |   StandardSGD s(0.0003, 1, 10000000000, -10);
  |                            ^~~~~~~~~~~
/<<PKGBUILDDIR>>/tests/callbacks_test.cpp: In function ‘void ____C_A_T_C_H____T_E_S_T____54()’:
/<<PKGBUILDDIR>>/tests/callbacks_test.cpp:473:28: warning: unsigned conversion from ‘long long int’ to ‘size_t’ {aka ‘unsigned int’} changes value from ‘10000000000’ to ‘1410065408’ [-Woverflow]
  473 |   StandardSGD s(0.0003, 1, 10000000000, -100, true);
  |                            ^~~~~~~~~~~

...

Test project /<<PKGBUILDDIR>>/obj-i686-linux-gnu
Start 1: ensmallen_tests
1/1 Test #1: ensmallen_tests ..................***Failed  5163.40 sec

The SGD maxIterations parameter to the SGD routine is stored in a
size_t. Arguably inappropriate; it belongs in something guaranteed to
be at least 64 bits, perhaps long long int.

Be that as it may, a test routine was passing a number meant to be so
large it would not be exceeded, to test a mechanism that should
terminate the optimization early. The large number chosen was
10000000000, ten billion, which exceeds the maximum value of an
unsigned 32-bit integer of 4 billion plus change.

This patch changes the maxIterations parameter passed to two billion,
2000000000, which is both sufficiently large for the purposes here and
safely below the maximum value for a 32-bit signed integer.

[rcurtin]

Looks good to me, thanks for the fix! We used to have some i386 build machines but don't anymore. I think it's fine to leave the maximum iterations as a size_t---really the goal of 10B iterations was just to be "really big but not unlimited". :)

[zoq]

Thanks, looks good to me.