Issue 150: Fix problems with encoding/decoding and add encoding unit tests.

[kaladay]

Resolves #150.

The main problem appears be a scoping issue mixed with an improper decoding of the URL.
The error message is a red herring that suggest an encoding problem when in fact it is a decoding problem.

This removes the decoding as well as manually percent encoding of white space.

This implements numerous unit test to help test out and prove this behavior.

This primarily focuses solving the problems with Fedora but many of the changes are broader than Fedora such that they may impact DSpace.

The data from the client should throw an error if the data is not properly escaped.
The escaping must not happen here.
This reverts the changes from commit eda72b9.

Simplify the getRdf() design.

• Replace logger.info() wrapped inside of logger.isDebugEnabled() with the simpler logger.debug().
• Extend the NotFoundException to add the exception to the message.
• Remove excessive logging.

The unit tests throw org.apache.jena.riot.RiotException errors to match what the server should be throwing when there is unencoded spaces.
It is unclear whether or not we should or should not convert this into a RestClientException.
This is left alone for further discussion.

The encodeSpaces() is entirely removed.
The project should not be encoding bad data.
Instead, bad data should throw an error, such as the RiotException.
The DSpace code will need to be reviewed for accuracy and regressions after this change.

Change fetchImageInfo() logic to handle RestClientException and return NotFoundException.

An observed potential problem is:

    private String getResourceId(String url) throws URISyntaxException {
        try {
            id = resourceResolver.lookup(url);
        } catch (NotFoundException e) {
            id = resourceResolver.create(url);
        }

This logic will fail on NotFoundException.
While this may not be the case here, there are a lot of places that translate the actual errors into NotFoundException even when the error is something other than NotFoundException.
This could potentially result in an attempt to create data when it should not.
I have not verified this behavior.

Several of the unit test regarding Fedora are refactored.
There are now multiple tests for encoded, decoded, and neither (uncoded).
The decoded tests are designed to throw exceptions that should be thrown under normal operations.

The Dspace unit tests will be updated in a subsequent commit.
Symbolic links are added for the json and rdf directories as a temporary measure to support the DSpace tests in their current form.

A bug with the processing of the percent encoded URLs is due to custom code being used to build the RDF Model.
Remove that code and use the Apache Jena API directly.
This fixes the problems but changes the network HTTP fetch approach that forces a change to the unit test logic.

Several unit tests are now failing, possibly due to problems exposed by the new testing approach.
I ran out of time and so I am simply disabling tests that I have not had enough time to determine what is wrong.

The unit tests exposed a NULL pointer error handling in the getMimeType() function.
The headers is being returned as NULL and it is not being safely handled.
The cause of this is unknown and needs further investigation.
Regardless of this problem, the getMimeType() should not be failing on a NULL pointer error and now no longer is.
The try..catch.. in that function is hiding the actual error and has been removed.

The original createRdfModel() has been copied over into individual tests to avoid rewriting all of the existing unit tests.
These needs to be changed and removed in the future.

[github-actions]

coverage: 48.862% (-33.2%) from 82.103%
when pulling a832e85 on 150-add_encoding_unit_tests
into 426d393 on cantaloupe_sprint_2_staging.