- fixed bug with unicode text formatting: the number of bytes

- fixed bug with unicode text formatting: the number of bytes
  per rune cannot be computed correctly if we have only parts
  of a rune - delay computation
- added html filtering mode: html tags and entities are ignored
  for width computations
- expanded tests:
  - extra tests for html text
  - extra tests that write text in various portions

R=r
DELTA=227  (126 added, 20 deleted, 81 changed)
OCL=20833
CL=20835