Enable finer granularity when defining CPU resources

[adwk67]

As a user I want to be able to define driver and executor cores without them being rounded up to an integral value of cores.

Notes

• This issue expanded the resource-usage documentation to make it clearer what the code does
• Cores are rounded up, probably because they were also used for spark.{driver|executor}.cores, which was removed in the above PR (as spark.kubernetes.{driver|executor}.request.cores - which we also set - takes precedence).
• This rounding should be removed to give more control to the user

[adwk67]

Duplicate: closing in favour of #363