For fixing error with labels for nvidia gpus/tpus being not
propagated/transmitted to environments properly

WIP feat: Changed resource config to allow custom values

* Fixed CI

* Fixed CI

* Fixed CI

* Catalog for e2e test adjusted to the latest starter

Now params which defined as `outer_namespace.inner_namespace.param_name` converted to the following structure before adding to pipeline:

```json
{
  "outer_namespace": {
      "inner_namespace": {
        "param_name": "value",
        ...
      },
      "another_inner_namespace": {
        "param_name": "value"
        ...
      }
  }
}
```

Added support for kedro namespaces for parameters

Now params which defined as `outer_namespace.inner_namespace.param_name` converted to the following structure before adding to pipeline:

```json
{
  "outer_namespace": {
      "inner_namespace": {
        "param_name": "value",
        ...
      },
      "another_inner_namespace": {
        "param_name": "value"
        ...
      }
  }
}
```

Co-authored-by: Dmitriy_Lamzin <dmitriy_lamzin@epam.com>