check zero buffer update, reformat pipeline layout, parallel shader loading

[ottoville]

• Fix unnecessary mapped buffer on buffer creation in certain cases
• Do not kill tab in chrome if updating buffer with 0 size data
• Use getBindGroupLayout to create a layout if necessary.
• Load fragment shader while vertex shader is being processing.

[ottoville]

I restored the _getPipelineLayout method into engine but also modified the _getBindGroupsToRender method to create a layout with getBindGroupLayout if layout has not been created.
The functionality has been described here gpuweb/gpuweb#446
This will make _getPipelineLayout method totally unnecessary, it can be deleted and still engine works. I did not delete it as I dont know how this will affect on performance.

[sebavan]

Love it !!! yup do not delete for now but will do in a bit I guess :-)