fix spec compliance

[hsitter]

as per https://www.w3.org/Submission/ws-addressing/ the only well-known
uri in 2004/08 was anonymous and it has a role prefix. from the
implicit reading I can find on earlier versions the same applies to them.

so, when constructing the well-known URIs inject a role prefix when
dealing with specs up to 2004/08 and raise a warning when attempting to
use anything but anonymous

[dfaure-kdab]

@caspermeijn Can you take a look?

[hsitter]

@dfaure-kdab btw, do you have an opinion WRT adding a Q_ASSERT(false) to the Anonymous condition? At first that looked like a good idea but then testing would be problematic so I opted not to assert anything.

[caspermeijn]

The difficulty is of-course that the four specs are subtly different. I think this change will make this difference between the specs more obvious.

[dfaure-kdab]

@dfaure-kdab btw, do you have an opinion WRT adding a Q_ASSERT(false) to the Anonymous condition? At first that looked like a good idea but then testing would be problematic so I opted not to assert anything.

Asserts prevent writing unittests for these conditions. So it might be better to not assert.
(Not to mention the risk that this gets triggered on data outside the control of the developer, such as server replies, or broken XSD/WSDL files)

[hsitter]

All comments should be dealt with now. I've also switched back to no assert and added back the test case for the bad-value-for-version scenario.

[caspermeijn]

Looks good to me

[hsitter]

ping

[hsitter]

actually in #193 for reasons...