[v0.9] Reduce BundleDeployment triggering based on deployed resources changes

[aruiz14]

Fleet agents have a mechanism to watch every individual resource that is deployed from a BundleDeployment. These watch events can be of multiple types, but the current implementation processes ADDED, MODIFIED and DELETED events in order to trigger a BundleDeployment reconciliation, which then should detect any drift between the desired and current state of all the deployed objects.
This operation can be expensive (although it's run on the agent), but since it updates the status conditions in the BundleStatus, it may results on triggering the Fleet controller in the upstream cluster as well. The idea of this issue is to revisit this implementation in order to reduce the frequency of those actions.

[manno]

Reverting because of CPU usage problems: #2258