Prevent int64 to int truncation in Shard API usage.

The function argument in `Shard` must be a function of two `int64` arguments. However, we are passing in a function with two `int` arguments. Thus, for large workloads, these arguments get truncated from positive `int64` values to negative `int` ones, resulting in a buffer out of bounds write.

PiperOrigin-RevId: 332557334
Change-Id: I236c9a2e7f53580e520571da8ba941a3aa9fa0b5

tensorflow/core/kernels/random_op.cc

@@ -202,7 +202,7 @@ class RandomGammaOp : public OpKernel {
     // avoid a couple flops which can be done on a per-alpha basis.
 
     auto DoWork = [samples_per_alpha, num_alphas, &rng, samples_flat,
-                   alpha_flat](int start_output, int limit_output) {
+                   alpha_flat](int64 start_output, int64 limit_output) {
       using Eigen::numext::exp;
       using Eigen::numext::log;
       using Eigen::numext::log1p;