The info is now fetched by the server on status change. This way the
info can be used in the notifications and also improves performance when
listing a huge amount of applications in the ui.

closes #349

With this change it is possible to include credentials in the instance
metadata which will then be used to access the client endpoints using
HTTP Basic authorization.

closes #359

With this commit you can now associate your applications with custom
metadata using `spring.boot.admin.client.metadata.*` sensitive values
are recognized by key and are masked in the http interface.

This can be useful in the future if there are instance specific settings
which should be used by the admin server, e.g. custom notification
recipients.

For now there is no extra imetadata view in the ui. The values are shown
in the environment view.

The StatusInfo has a new details field. The StatusUpdater puts either the
reponse from the health endpoint or the exception message in the details.
This allows us to see the details over time (via the journal) and also to
use `em in a bit more mail notifications.

closes #329

With this commits the server no longer depends on the client. The
classes for the clients are moved to de.codecentric.boot.admin.client
package.
Since the perspective on applications is totally different from client
and server the model has been split apart.

What means this for users:
1) The client is no longer a transitive dependency for your admin server
you may need to add it explicitly to your dependencies, respectively
jolokia-core in case you're using discovery and want to use the jmx-bean
interface

2)In case you have cutomised parts of the client you need to reflect the
package change.

fixes #332

With the fix for #260 we ignored all events from child contexts which turns out
to be wrong since the "main" application context is, in case spring cloud config
is used, is a child  of the bootstrap context. So we just act on the events from
WebApplicationContexts. After they're fired up the needed port information is
available.

fixes #277

When includeing spring-cloud-netflix-hystrix-amqp to your application it looks
like that the ApplicationReadyEvent is fired twice (from main and child
application context). So we need to take care that only the events from the
main applications are taken into account.

fixes #260

As it turns out (rather obvious) that the ThreadPools for registration and
status updating needs to be destroyed on application shutdown so that the
worker threads gets cancelled. To achieve this we simply register the
ThreadPools as spring beans (since they are implementing DisposableBean).

fixes #253

This adds a new FilteringNotifier which can use filters to suppres certain
events from being notified of. Two filter implementation are provided, one
filtering by name and one by id. Additional when a FilteringNotifier bean is
present, the autoconfiguration creates a REST-controller to add and remove
filters at runtime.

 * Use @EventListener instead of ApplicationListener-interface
 * Don't inherit ClientApplicationEvent from ApplicationEvent - so JournaledEvent isn't needed anymore
 * added eclipse formatter rules
 * reformatted all files

fixes #105

The health-check is made at a fixed interval (default: 10s) for all expired status-information (default-lifetime: 30s).
The application health-endpoint must either respond with a JSON-Body having a status-field or with an empty body and all 2xx-http-responses are interpreted as UP and every other status as DOWN.

Rename ApplicationStore methods to spring-data-style

* Separate AppStore from AppRegistry
* Allow Id generation to be altered
* Remove logic from Application; Move id genretation into AppRegistry
* Make AppStore threadSafe