lynx-dev chartrans broken with SOURCE_CHACHE

[Klaus Weide]

1. Display character set (as saved in .lynxrc is 'Western (ISO-8859-1)'.
2. Start lynx as

     lynx http://www.cogsci.ed.ac.uk/~richard/unicode-sample.html

3. Check that things look right, especially in the 'Latin-1 Supplement'
   block.

4. Go to 'O'ptions screen, change display character set to
   '7 bit approximations (US-ASCII)'.   (I used the old style Options
   Menu, in case it matters.)

5. Look at 'Latin-1 Supplement' block.
   a. If SOURCE_CACHE was none, everything is ok.
   b. If SOURCE_CACHE was FILE or MEMORY, 8-bit characters are still shown
      directly.

Best tested with a slang-compiled lynx, it shows C1 characters (illegal
in the now-current display character set) in slang's escape
representation.  (With a curses-compiled lynx they should be filtered out
when displaying).  One can also use 'P'rint, with a printer definition
that invokes less (or similar) *without* a LESSCHARSET or -r, to see
what's in the HText structure.

Modifications:

1. isn't strictly necessary.  The same applies if display character set
is for example KOI8-R, just look at the appropriate block then.

In 4. try one of the RFC 1345 choices instead of '7 bit approximations
(US-ASCII)'.

Leonid, can you look into this?

  Klaus