Add an --all-namespaces option to tail all pods across all namespaces

https://rust-lang.github.io/async-book/07_workarounds/03_err_in_async_blocks.html

We need to handle errors correctly in async blocks. My quick pass on it i was getting
Box<dyn ::std::error::Error> can not be sent between threads safely

Will need to spend some more time on this

Need to add a Containers CLI Arg that accepts multiples args. That we we can tail logs of specific containers instead of all of them.
--container foo --container bar
-c foo -c bar

https://github.com/clux/kube-rs
The one main feature I need is following logs which isn't implemented yet. Trackable at
kube-rs/kube#109

This would also allow me to write an informer, that can pull in new pods as they spin up.

Exmample:
https://github.com/clux/kube-rs/blob/master/examples/event_informer.rs

We need to update the informer to see if a label or such has been added or modified during an edit, so we can pull it into the log streaming pool if theres a --selector argument.
1.) In the match statement under modified, see what we can key off of if a deployment/pod hais edited
2.) look up the pod that was edited, see if label matched what --selector key/value pair was passed in, if it was add to pool.

An outffile flag needs to be added to the cli to change the default path

Only return logs newer than a relative duration like 5s, 2m, or 3h

Code could be simplified now that there is an outfile argument that defaults. Could probably do away with the generate_hashmap method. Also could probably swap to using actual rust kube bindings, and take advantage of the watcher API so we can update the logs if new pods are deployed.

in the threads, we spin off system processes. We need to make sure these are being cleaned up when signals are being sent to kill the program.

Write now wufei uses system threads. Maybe good to try to test out other methods of concurrency (futures, async, green threads via 3rd party lib)

I tried to call pods before run_individual and pass it in as a reference beforehand when collecting futures/tasks. I tried to hack around it and kept getting errors

   --> src/kube.rs:121:53
    |
117 |       let current_pods = Api::v1Pod(KubeClient::client().client.clone())
    |           ------------ move occurs because `current_pods` has type `kube::api::typed::Api<kube::api::resource::Object<k8s_openapi::v1_15::api::core::v1::pod_spec::PodSpec, k8s_openapi::v1_15::api::core::v1::pod_status::PodStatus>>`, which does not implement the `Copy` trait
...
121 |           children.push(tokio::task::spawn(async move {
    |  _____________________________________________________^
122 | |             run_individual(&pod, current_pods.clone()).await.unwrap()
    | |                                  ------------ use occurs due to use in generator
123 | |         }));
    | |_________^ value moved here, in previous iteration of loop

If selector label is added, pod name is ignored. Get all pods in a group

support more the one json key

Have base functionality for writing to a file already included. Keeping the file open and writing the logs until CTL + C is pressed

Right now we just grab the logs for the service container. May need to use the istio-proxy for each service (this could be useful in a/b deploy situations and such)

Using unwraps everywhere. Need proper error handling.

Add a json parser using serde to record messages that match json strings/keys. This will allow for easy searching if you are looking for something in logs.