This PR aims to HEAVILY improve performance by running the "build" portion of the prompt only when necessary.

Before, the script checked for colors, shell, and many commands on every prompt. Now, we build the prompt only after it is updated.

Now, a few things about this PR:

1. I have NOT tested for ZSH, at all. I don't use ZSH, so I wouldn't even know how to validate it.
2. I did test for bash as much as I could, and I noticed no problems, but I'm sure there's some stupid edge cases I haven't accounted for.

Overall, I didn't find a way to "benchmark" this. I had been running a personal, heavily modified version of this script, which only accounts for Bash, Linux, tput being available, etc, and while doing that and building the prompt, I get a 15x (yes 1500%) speed improvement, which I benchmarked like this: time { for i in {1..1000}; do kube_ps1 > /dev/null; done }.

However, this doesn't work for this script since we do it "lazily". I'd appreciate any way to benchmark this. I'd also appreciate some validation on ZSH before I mark this PR ready to review.