Read HttpServer2 resources directly from the source tree (if exists)

Currently the Hadoop server components can't be started from IDE during the development. There are two reasons for that:

1. some artifacts are in provided scope which are definitelly needed to run the server (see HDFS-12197)

2. The src/main/webapp dir should be on the classpath (but not).

In this issue I suggest to fix the second issue by reading the web resources (html and css files) directly from the source tree and not from the classpath but ONLY if the src/main/webapp dir exists. Similar approach exists in different projects (eg. in Spark).

WIth this patch the web development of the web interfaces are significant easier as the result could be checked immediatelly with a running severt (without rebuild/restart). I used this patch during the development of the Ozone web interfaces.

As the original behaviour of the resource location has not been change if "src/main/webapp" doesn't exist, I think it's quite safe. And the method is called only once during the creation of the HttpServer2 there is also no change in performance.