Document scheduling recommendations

[LucasMrqes]

I suggest we add a dedicated section to the documentation with guidelines and best practices for configuring Burrito pod scheduling, including affinity and nodepool strategies
Fine-tuning the scheduling of some components may be crucial to achieve the best performance at a minimal cost with Burrito.
A few examples:

• At least one instance of the datastore and hermitcrab should run on each availability zone, to minimize inter-zone traffic.
• To reduce inter-node traffic to the maximum, an option is to deploy Hermitcrab as a daemonset on a dedicated nodepool for Burrito runners