Provide new mechanism for injected container configuration

Nobody has claimed this yet.

## Feature Request

### What problem are you trying to solve?

As the Kubernetes container spec evolves, there will be new options for configuring the linkerd-proxy and linkerd-init containers that are added when running `linkerd inject`. Rather than adding new CLI flags for for every possible container configuration, we need a better way for users to modify the injected YAML. This will allow them to configure things like resource requests/ resource limits / readiness probes / liveliness probes without an explosion of CLI flags.

### How should the problem be solved?

From the comments on #2001, it seems like we could instruct users to use [kustomize](https://github.com/kubernetes-sigs/kustomize) or some similar tool to modify the YAML post-inject.

### Any alternatives you've considered?

A CLI flag for every field in the [Kubernetes container spec](https://kubernetes.io/docs/reference/generated/kubernetes-api/v1.13/#container-v1-core) 🙀

Start with the `linkerd inject` entry point and review the discussion in #2001, along with the Kubernetes container spec referenced in the issue. Determine how users could modify injected YAML for proxy and init-container settings without adding a flag for every field. Done should provide a supported mechanism for those post-injection changes.

Written by the indexing model from the issue text.