Use IDs for attachment URLs

In Confluence 2.2, we migrated to using IDs for attachments on disk, in order to solve problems with the character encoding of filenames on some operating systems.

We see many problems with attachment filenames in URLs, usually due to Tomcat's URL encoding setting. We should use attachment IDs in the URLs for attachments in Confluence. This would automatically solve many of the problems.

The only potential issue is around browsers that might use the URL in order to provide a filename when downloading an attachment. Some browsers might also sniff the URL to determine the content-type, resulting in different ways of displaying a document. We need to test these, and work out a suitable workaround.