docs: document our branching / versioning process

We want to be more systematic about our branching / versioning.

Document the process we were (trying to) follow.  We can also
automate some of the toil here.

Author: justinsb

docs/project-management/branching-and-versioning.md

@@ -0,0 +1,35 @@
+# Branching and versioning strategy
+
+## Versioning
+
+kubebuilder-declarative-pattern is a library, that works with controller-runtime and kubernetes.
+
+We follow semantic versioning, similar to other tooling in the kubernetes ecosystem.
+
+Because both controller-runtime and client-go introduce changes that require code changes, we follow their versioning.
+Specifically we align with the controller-runtime version, which itself aligns with kubernetes versioning.
+
+Thus:
+
+| kdp version | controller-runtime version | client-go version |
+|---|---|---|
+| v0.20 | v0.20 | v0.32 |
+| v0.19 | v0.19 | v0.31 |
+| v0.18 | v0.18 | v0.30 |
+| ... | ... | ... | 
+| v0.x | v0.x | v0.x+12 |
+
+
+TODO: Should we align with client-go version?
+
+## Branches
+
+We maintain a `release-0.x` branch for the kdp major version `0.x`.  Patch versions are cherry-picked to the branch and released.
+
+We cut the release branch with the beta release (i.e. we create `release-1.100` when we tag `1.100.0-beta.1`).
+
+Before beta releases, the master branch is for the next minor version.  If we tag a version, it would be (for example) `1.101.0-alpha.1` (and then `1.101.0-alpha.2` etc), assuming the most recent release branch was `1.100`
+
+TODO: Should we rename master to main now?
+
+TODO: Do we want to do a bulk catch up, where we create releases aligned with each version we missed?