Switch to JAXRS server implementation

[lukasgraf]

This PR changes the client implementation in ftw.tika to talk to a Tika JAXRS server (HTTP / RESTful) instead of the old builtin server from tika-app.jar (plain TCP).

The old TCP server is not supported by ftw.tika >= 2.0 anymore.

Changes

tika.cfg example / testing buildout

• the bin/tika-server script now starts a Tika JAX-RS server (tika-server.jar), not the builtin CLI server from tika-app.jar anymore
• The new server now listens on port 9998 by default
• Buildout sections in the example buildout are now named more precisely (tika-app-download and tika-server-download instead of just tika-download).
• The *.jar files are now downloaded from http://repo1.maven.org (archive.apache.org doesn't mirror the tika-server.jar)
• The Tika version used in the example buildout was bumped to 1.5

  Implementation

• The client implementation in ftw.tika now expects to talk to a Tika JAXRS HTTP server
• Word bookmarks are stripped from returned plaintext

@phgross @maethu please review

/cc @jone

[lukasgraf]

Tests are flaky right now, working on it. I may have to increase the delay for the testrunner to wait for the Tika server to boot up.

[lukasgraf]

@phgross @maethu fixed flaky tests.

[phgross]

Just the detail I've commented above, otherwise looks :+1:

[jone]

@lukasgraf nice! I think we should use another port for the JAXRS implementation in the ftw-buildouts, maybe with a new buildout config file to extend from. When upgrading tika, the buildout URL can be modified. The existing tika.cfg in ftw-buildouts should not be changed to JAXRS so that a rerun with old tika pinnings still works. What do you think?