- Move scion token strings to their own interface.
- Simplify tab-to-space conversion and use provided Eclipse
  infrastructure
- More code template hacking.

- Take the Unicode approach to lexing identifiers.
- Cosmetic changes to HaskellEditor
- Take the Unicode approach to lexing groups of symbols.

- Write scion server output to its IOConsole in 1K chunks to improve
  UI responsiveness (could probably use a lower threshold, but 1K
  is what JP used to use.)
- Add documentation to ThingAtPointCommand, update formatting.
- Remove unused class NOPContinuation.

- Remove unused class ConsoleCompilingReporter.
- Add high/low water marks so that IOConsole doesn't cause Java
  OOM errors.

Add initial support for lexer tokens returned by GHC and the
scion-server "tokens" command. These will be used by the UI's
content assist to get the previous token in front of the editor's
point to give more context to the completion assistant.