[TOPI] Fix data race of batch multibox detection

[wrongtest-intellif]

• For cpu, currently the buffer space of box_coord and pred_coord is shared across parallel batch like

for i in parallel(batch_size):
    # ...
                for l in range(4):
                    box_coord[l] = anchor[0, j, l]
                    pred_coord[l] = loc_pred[i, j * 4 + l]

• For cuda, fix two index usage error of batch idx.

[tvm-bot]

Thanks for contributing to TVM! Please refer to the contributing guidelines https://tvm.apache.org/docs/contribute/ for useful information and tips. Please request code reviews from Reviewers by @-ing them in a comment.

• No users to tag found in teams: topi _{See #10317 for details}

_{Generated by tvm-bot}

[wrongtest-intellif]

@tvm-bot rerun

[wrongtest-intellif]

cc @kevinthesun @Laurawly @tqchen Hi~ could you kindly take a look at it, thanks!

[tqchen]

cc @masahi , can you help to take a look at this?

[masahi]

It seems box_coord is the same in all threads. How about passing anchor, j, and l to _hybridy_transform_loc instead?

[wrongtest-intellif]

good idea~ updated as suggested.

[masahi]

Need to fix lint.

[masahi]

@tvm-bot rerun

[wrongtest-intellif]

@tvm-bot rerun