The introspection functions are new API.

For conversion functions, we use the name `convert`.
So let's be consistent with introspection functions.
It's also more concise.

The submodule is called versions; let's name the test file consistently.

To me, the name `find_java` suggests we will be locating a JVM
installation, rather than "finding" information about Java objects.
The information doesn't need to be "found" or "located", but rather
only introspected or interrogated. Technically, I suppose
"introspection" implies read/access while "reflection" implies
write/mutation, but `jintrospect` is rather clunky, whereas the
term "reflection" is widely known in both Java and Python circles.

For consistency with scripting integration tests.

More consistent with the rest of the library.

* scyjava.fields -> scyjava.inspect.fields
* scyjava.methods -> scyjava.inspect.methods
* scyjava.src -> scyjava.inspect.src

And add new `constructors` and `members` convenience functions.

This code worked:

  import scyjava
  print(scyjava.config)

But this code didn't:

  import scyjava
  print(scyjava.inspect)

Because scyjava.config was being imported in another file
further down the chain. Better to be explicit about wanting
both of these submodules available at the top level.

And do not try so hard with exception handling; it should be up to
higher level functions like scyjava.inspect.src to catch such failures.

When invoking scyjava.inspect functions, they can optionally report the
source URL at the top, before printing the members. But this only works
if scijava-search is on the classpath. Let's let the source flag default
to None, in which case it swallows source code URL detection failures
gracefully, to make the common case of scijava-search not being
available work without hassle.

And let's have the various inspect functions accept static and source
boolean flags, which get passed along to the internal _print_data
routine, as was previously the case when they were partial functions.