Fully enforce reusability in MP Metrics 1.1. support

[tjquinno]

The code that supports MP Metrics 1.1 does not fully enforce reusability. These changes correct that.

(This PR also addresses some unnecessary references to the internal bridging infrastructure which these classes did not need to use; they can and should depend directly on the MP metrics 1.1 artifacts.)

See issue #1046

[tjquinno]

The (tentative) changes to FT to make the automatically-created metrics reusable are because automatically-derived metric names come from the class name and method name of the annotated method. In the FT unit tests, the CircuitBreakerBean uses FT on two overloaded variants of exerciseBreaker. Both methods yield the same automatic metric name, so if the corresponding metric is not set up as reusable then, now that the registry code is enforcing the reusability, FT tries to reregister the same metric twice which fails.

[tjquinno]

Note that the preceding changes to MetricsSupport register the vendor metrics as reusable. This is needed because the TCK starts the server multiple times, which causes the MP services to be loaded multiple times, but there is no life cycle support in the server for services to respond to server shutdown so they can clean up. With the fixes to reusability enforcement, and with reusability being false by default in metrics 1.1, this caused failures in the TCK as the vendor metrics were reregistered. Specifying these metrics as reusable in 1.1 is consistent with our 2.0 implementation in which programmatically-created metrics (as these are) are reusable by default.