fix out-of-bounds read in vk_string_validate on truncated utf-8

[aizu-m]

Found while running the loaders layer/extension name validation under ASAN with malformed UTF-8.

1. on a multi-byte lead byte vk_string_validate sets num_char_bytes and then scans that many continuation bytes.
2. the inner scan stops at max_length but never at the strings null terminator.
3. a requested layer/extension name ending in a lone lead byte (eg 0xC2/0xE2/0xF0) makes it walk past the terminator, and the outer loop then resumes past it, so ASAN reports a heap read up to MaxLoaderStringLength (256) bytes past the buffer.
4. stop at a null seen inside a sequence, mark it bad data and return; added a regression test that trips the old read under ASAN.

[ci-tester-lunarg]

Author aizu-m not on autobuild list. Waiting for curator authorization before starting CI build.

[ci-tester-lunarg]

Author aizu-m not on autobuild list. Waiting for curator authorization before starting CI build.

[ci-tester-lunarg]

CI Vulkan-Loader build queued with queue ID 16340.

[ci-tester-lunarg]

CI Vulkan-Loader build # 3584 running.

[ci-tester-lunarg]

CI Vulkan-Loader build # 3584 passed.

[charles-lunarg]

There should be a better way to do unit level tests. But for the time being, just exporting the function directly is simple enough to get the job done.