TblSpark inheriting from LazyTbl, fix LazyTbl's name

[machow]

Let's try the dbplyr approach, which...

• treats it largely as a sql source, using the default translator
• allows sparklyr to implement custom S3 methods for copy_to and collect

TODO:

Will need to modify singledispatch for...

• collect - to execute on the spark source (using pyspark)
• copy_to - (also need initial implementation), creates a TblSpark

For the tests we'll need to extend conftest.py::Backend so we can load dataframes willy nilly.