Clarify the behavior of continuingRequest()

It's not exactly true to say that ThreadLocalRequestScopePropagator
doesn't propagate *any* request-specific information, since the
context implementation type typically does contain some request-
specific information, which can in turn be used to derive the
necessary RequestScoped objects in the new thread.

Clarify this behavior in the javadoc, and give an example. This is a
fairly technical description, but the target audience is someone
implementing a new RequestScopePropagator, so they already need to
understand the request scope internals pretty well.

Change-Id: I3326444c76d305fa4b067288ab2a3c7e4ad32997

gerrit-server/src/main/java/com/google/gerrit/server/util/ThreadLocalRequestScopePropagator.java

@@ -74,6 +74,13 @@ public abstract class ThreadLocalRequestScopePropagator<C>
   /**
    * Returns a new context object based on the passed in context that has no
    * request scoped objects initialized.
+   * <p>
+   * Note that some code paths expect request-scoped objects like
+   * {@code CurrentUser} to be constructible starting from just the context
+   * object returned by this method. For example, in the SSH scope, the context
+   * includes the {@code SshSession}, which is used by
+   * {@code SshCurrentUserProvider} to construct a new {@code CurrentUser} in
+   * the new thread.
    *
    * @param ctx the context to continue.
    * @return a new context.