feature(kafka): design for testing scylla connectors

docs/kafka-setup.plantuml

@@ -0,0 +1,31 @@
+@startuml
+skinparam nodesep 30
+skinparam ranksep 30
+
+
+legend
+|= |= API |
+|<back:red>   </back>| CQL |
+|<back:green>   </back>| Kafka API |
+endlegend
+
+database "DB Nodes" {
+    card [source table]
+}
+
+package "Loaders" {
+    card [cassandra-stress]
+    card [kafka-consumer]
+}
+
+package "Kafka" {
+    card [kafka]
+    card [kafka-connect]
+}
+
+
+[cassandra-stress] -[bold,#red]-> [source table]: **1)** c-s write\n
+[kafka-connect] -[bold,#red]->  [source table]: **2)** Kafka connector read CDC
+[kafka-consumer] -[bold,#green]->  [kafka]: **3)** reading from\nkafka for validation
+
+@enduml

docs/kafka.md

@@ -0,0 +1,53 @@
+# Kafka support in SCT
+
+## Requirements
+
+1) Be able to run a kafka setup, with a db cluster SCT created
+2) Have any of scylla connectors installed on it
+   1) from a tarball (local file / URL)
+   2) from confluent-hub (latest/specific version)
+3) Work locally with docker backend
+
+## Design
+
+![Design](./kafka-setup.png?raw=true )
+
+### Local setup
+
+we should be working with a https://github.com/conduktor/kafka-stack-docker-compose
+and it would run where the test is running (on the SCT runner)
+for case it won't be enough, we should support running it on it's own machine
+
+### High load setup
+
+for high load setups we can go in few routes
+* **SaaS** - confluent cloud / AWS MSK / others
+  - Pros:
+    * All configuration and setup, and would match better how user would be using scylla connectors
+    * AMS MSK, we are already working mostly in AWS, and familiar with operating its services
+  - Cons:
+    * Might be pricey
+    * Connectivity to our test setup might be a bit more complex
+    * on MSK we'll need to run your own kafka-connect instances
+    * there are limitations to install custom connectors on SaaS
+
+* **VMs** - create and configured our own VMs and cluster
+  - Pros:
+    * We will have full control how it would be setup
+    * it's fit the current model of SCT is built (or supposed to be)
+  - Cons:
+    * We will need to know all the different ways to configure and optimize kafka deployment
+    * We don't want to be kafka "Experts" to that level...
+    * We will need to keep it updated all the time
+* **K8S** - deploy using helm/operators onto k8s cluster
+  -Pros:
+    * We have all the code to support k8s setup on kind/EKS/GKE
+    * Operation and setup into k8s should be relativity easy, and designed to scale
+    * it would fit testing with scylla-operator
+  - Cons:
+    * manage setups with k8s isn't that easy, lots of "buttons" to play with
+
+## Open questions
+
+* how validation is going to be done ? with `kafka-console-consumer` ?
+  with python code we'll need to write https://kafka-python.readthedocs.io/en/master/apidoc/KafkaConsumer.html ?