Consistently resolve path of jruby.bash

[mrnoname1000]

The bash script sometimes returns an incorrect path for JRUBY_HOME. I came up with a bunch of cases where it would return the wrong directory, in no particular order:

1. a directory in PATH is a symlink
2. a directory in PATH is in a symlink
3. the executable is a symlink in PATH
4. the executable is in a directory that is a symlink
5. the executable is a symlink that points to an executable in a directory that is a symlink
6. the executable is being read from a pipe

Simply replacing $0 with $BASH_SOURCE fixes case 3. Obviously this is bash-specific, so this can't be ported to the sh script.
Case 4 are solved by using readlink on the executable until it isn't a symlink. The script already did this, so I made some changes to solve other issues.
Cases 1, 2, and 5 are solved by running pwd -P in the directory of the executable.
I don't know how to solve case 6 without hardcoding a directory to check.

Because pwd -P always returns the absolute working directory, it seems running dirname "$SELF_PATH" is unnecessary, so I removed it.
If anyone can think of any other torture tests, let me know.

[headius]

Looks good, though I've never seen that syntax you used for JRUBY_HOME. What is that?

Thanks for the patch!