This commit switches to using the new Travis container-based
infrastructure.

This commit makes Travis launch sauce connect, instead of
using karma to start Sauce Connect

Make Util.getPosition be relative to page

Commit 5108c463 caused a regression
in the case where scrolling is used -- getPosition return position
relative to the viewport, while getEventPosition expected a position
relative to the page.

As per
https://developer.mozilla.org/en-US/docs/Web/API/Element/getBoundingClientRect,
the fix for this is simply to add the `pageXOffset` and `pageYOffset` to
the output of `getBoundingClientRect()`.

Fixes #459.

This commit increases the browserNoActivityTimeout to 100s
(from the default of 10s) in an attempt to alliviate the
timeout issues that we are seeing.

This commit updates the test dependencies to the latest versions.

Fix disconnect/reconnect issues

Commit 155d78b3 prevented reconnections
from working properly.  This fixes that by creating a new RFB object
after disconnecting or failing.

Furthermore, this ensures that a new connection cannot be opened util
we've actually disconnected (either by timer or by receiving a `close`
event).

Closes #452

This commit removes the old debian packaging information, which is
out of date.  People who wish to create Debain packages for noVNC
are better off using the package information from the actual Debian
or Ubuntu packages.

Closes #453

Issue #449 pointed out that there were some files that were
missed in 6f514864.  This
fixes that.

Closes #449
Closes #450

When `utils/launch.sh` clones websockify, it can be cloned
into the incorrect directory, depending on how `utils/launch.sh`
is run.  This commit ensures that websockify is always cloned into
`utils/websockify`.

Somehow, `utils/websockify` itself manage to sneak back in to
6f514864.  This actually removes
it.

Remove local copies of websockify

Replace custom getPosition algorithms with getBoundingClientRect.

Closes #434

The current noVNC release is 0.5.1, but package.json still
says version 0.5.0.  This fixes that.

Fixes #441

This commit removes local copies of websockify.
Instead `utils/launch.sh` performs the following logic:

If `utils/websockify` exists, use `utils/websockify/run` (if the latter
does not exist, or is not executable, fail, since this is probably a
mistake).

Otherwise, check to see if websockify is installed somewhere (try
`which websockify`).  If it is, use that.  Otherwise, clone
websockify from github, and tell git to ignore that directory.

Packaged versions of noVNC should simply list websockify as a
requirement.  The debian packaging has been updated to reflect
this.

Closes #433

Support ExtendedDesktopSize and SetDesktopSize

* Added new tests for the setDesktopSize encoding
* Added new tests for the ExtendedDesktopSize encoding

* Added a resize request (setDesktopSize) triggered when connecting and by
  changes to the browser window's size.
* Hid the view-drag-hand when the display area is the same or smaller than the
  remote session size.
* Added a setting for the automatic resize feature.
* Updated vnc.html and vnc_auto.html to reflect the changes to the UI.

keysym.js needs inclusion in debian's install list

* Support sending the setDesktopSize encoding (client -> server)
* Support recieving the ExtendedDesktopSize encoding (server <- client)

* Split viewportChange into two functions, one for changing size and the other for changing position.
* Modified viewport code to be capable of changing to a bigger size in the context of a
  client-initiated resize.
* Made clearer distinctions between when viewport-clipping or not.
* Added public function for telling when viewport-clipping.
* Updated tests that were using viewportChange.

Unregister event listeners from websock.

Prevents possible memory and event notification leaks when tearing down
connection and reestablishing a new one.

Loads keysym.js

Without it, `XK_ISO_Level3_Shift` (at least) is undefined.

Bonus 3 bug fixes:
 * Meta is 0xFFE7 not 0xFE07
 * Super_L is 0xFFEB not 0xFFEC
 * Super_R is 0xFFEC not 0xFFED

The HEXTILE test which tested for a background tile followed
by an empty tile was only wide enough to actually test for one
tile, thus not actually testing the functionality.  It now actually
uses two tiles, thus actually testing the functionality.

Fix for typo from a recent pull

There is a minor typo in the recent pull  Simple fix.

Closes #412

Previously, if a HEXTILE tiles was received with a subencoding
of 0x00, it would draw a rectangle using data from the render
queue, which would result in random colored blocks when using
the HEXTILE encoding.  This is the result of a miscopy during
the refactoring.  It now has the correct functionality according
to the RFB protocol specification, which is to draw a rectangle
with the last set background color.

Closes #411