Python API documentation generation tool Code

- Added support for the syntax `name <target>` in xlink.

- If the source file is not encoded correctly, then issue an error
(rather than crashing) and give up on parsing.

- Minor css fix for subclass list

- If the source file is not encoded correctly, then recover as well
as we can, and issue a warning.

Fixed a bug that caused exceptions to propagate when reading a
metadata variable failed.

- Moved TerminalController to epydoc.util
- Added TerminalController.render

- Removed restriction that --text can only take one name
- Fixed bug that caused --text and --check to fail
- Put blank lines around generated text docs

- Added a few regression tests for the command-line interface

- epydoc will now accept multiple action arguments (eg --html, --pdf)
- if the target option (-o) is used before any action arguments, it
gives the default target name. If it is used after any action
argument, then it applies just to the action argument that preceeds
it. Note that this is backwards compatible with the old case where
only one action argument could be used -- either way, the target
argument would apply to that one action argument.
- In config files, the "output" option can now be used with multiple
output formats. To specify per-format output targets, use the
form "format->target" e.g. "html->myhtmldir" or "pdf->foo.pdf".
The "target" option sets the default output target.
- if latex output is not explicitly requested, then latex output is
now written to a temp dir, and cleaned up after execution. The target
name for dvi/ps/pdf outputs is treated as a filename.
- parse_arguments now just returns the options argument -- returning
a tuple with the names was redundant since we store the names as
options.names anyway.

- Fixed spurious warning about identifiers occcuring in two groups
when pre-2.4-style decorators are used.