Warn on missing testdata submodule

[SimonSapin]

Steps to reproduce:

git clone https://github.com/html5lib/html5lib-python.git
cd html5lib-python
tox

The tests fail with error messages like:

FileNotFoundError: [Errno 2] No such file or directory: '…/html5lib-python/html5lib/tests/testdata/encoding/chardet/test_big5.txt'

The reason is that git clone does not clone/checkout submodules by default. The fix is to run git submodule update --init.

When finding html5lib/tests/testdata empty, the test harness should print an appropriate warning with instructions.