Add NodeFeatureRules to label Nvidia NICs

[wmousa]

This patch does the following:

• Add "Nvidia NICs PCI" NFR to lable the node with pci-15b3.present": "true" if any of Nvidia NICs is exist
• Add "Nvidia NICs SRIOV" NFR to lable the node with pci-15b3.sriov.capable": "true" if any of Nvidia NICs has SRIOV enabled
• Add nfd.nodefeaturerules flag and set it true by default to allow creating NFRs above
• Remove the networking PCI class from the source configuration

[ArangoGutierrez]

I would leave this, in case the users decide to set deployNodeFeatureRules to false. NFD can handle the duplicated config giving NodeFeatureRule priority

[adrianchiris]

I suggested to waleed to remove to avoid having duplicate places that label the same thing.

if this is left here, then why create nodefeaturerule by default. i think we want to rely on one mechanism for most cases.

since we may have external NFD deployed (which we require to have nodefeaturerules enabled) which then will require us to use nodefeaturerules created. i prefer to always rely on it.

if user decides to set this one to false, he needs to be aware of the implications. IMO.
in that case the user either:

1. has another NFD deployed with all needed configs
2. user would need to change this entry to add the values needed
3. or simply keep deployNodeFeatureRules true

[wmousa]

ACK, Done

[adrianchiris]

we need to add a section in README on the different ways NFD can be deployed.

1. externally
2. via network operator

if externally, we need to specify requirements. i.e version, expected features to be enabled and install example using helm from upstream repo