Minimalize the transitive dependencies pulled in by tika-parser

We use tika-parser (1.3) which has a very large footprint of 'optional' dependencies, many of which we don't need and shouldn't by default pull in.

Default excluded tika-parser parsers are:

• PKCS7 signed messages (bouncycastle bcmail/bcprov)
• audio and video formats (vorbis, mp4)
• NetCDF and HDF
• MIME4J (raw email and mbox files)
• EXIF (image full text indexing)
• Rome (RSS and Atom feeds)
• Common Compress (archives like zips) (this had to be reverted because of regression in CMS7-9412)
• asm (Java classes)
• Boilerpipe (surplus "clutter" around main html content)

Anyone needing such resources indexed with tika-parser can and should add the needed dependencies explicity instead.