[core] Fixes for drupal_strpos()

Merged #276453: Tests needed: unicode.inc tests into this one.

The tests revealed three issues in drupal_strpos(), two of them were already documented:

• #319310: Off-by-one error in drupal_substr() results in "undefined index" error in includes/unicode.inc line 516: there were off-by-one errors at several places of the emulated implementation of drupal_substr(). I restored the symmetry of the function, and fixed the end-of-loop condition. That solved the issue.
• #340310: drupal_strpos vs native strpos inconsistency: drupal_substr() missed the $length = 0 case.
• Closely related to the first issue, drupal_substr() returned wrong results when $length > strlen($string)

@Dries: #212130: decode_entities() doesn't support all entities is a valid issue. The entity decode test (mostly written by Charlie in #276453: Tests needed: unicode.inc), will be easily extended by the additional entities we should manage.

#212130: decode_entities() doesn't support all entities now has a proper patch (and test extension), that depends on this one.

To clarify:

- This patch introduce a test for decode_entities (written by Charlie in #276453: Tests needed: unicode.inc). The test doesn't cover any of the missing entities so it pass correctly
- The patch in #212130: decode_entities() doesn't support all entities extends the tests of that patch and adds the missing entities (along with very significant performance and memory usage enhancements).

Quick code style followup for 7.x, then we can mark to for 6.x.

No, please don't. This non-standard indenting is the only way to make sense out of the test case when reading it.

@Dave: I think it makes sense to keep it that way, but I won't battle over it. Feel free to re-bump to D7 if you really don't like it.