Probably a good idea because of the reverted Event-B EMF dependencies.
This also matches what I wrote in the release history.

One less thing to bump manually when making a new version.

Tycho seems to use the .classpath in some way, because it prints stack
traces if there are nonexistant entries in it. Apparently the outdated
.classpath hasn't caused any actual problems so far, but let's fix it
just to be safe.

Apparently this isn't a new bug. But at least in the case of our
ABZ 2023 AMAN model, it only occurs with Event-B EMF 7.x (core 6.x) and
not with 6.x (core 5.x). So reverting this dependency is an improvement,
I guess?

We don't care about generic EMF objects, only about EventBElement
(right..?). This noticeably improves performance when saving (almost
twice as fast for a medium-sized machine with 30 invariants).

This way they are visible to normal users.

This way, we don't have to explicitly list all EMF-EventB core package
names, and avoid potential problems with other plugins that use the same
unqualified names.

Modifying an already existing variant worked, but when adding a new
variant to a machine that didn't have one, Rodin complained that the
variant has an empty label.

This fix requires increasing the minimum Event-B EMF version to
guarantee that Variant.setName exists.

The description text area in the Eclipse plugin install dialog is only
three lines tall by default, so this makes the newest entry visible
without scrolling.

To prevent somehow installing old parsers together with new Camille.

Despite the major version bump, the EMF model and Java API *seem* to be
backwards-compatible, at least for our purposes.

This fixes the version conflict that prevented installing Camille
together with current versions of CamilleX.

To avoid installing incompatible major updates.

This is impractical to maintain. It looks like no other Rodin plugins
specify these directives, so I'm assuming it's safe to remove them.

Contains *less* than camille.target and also seems to be older.