Update and test the getting-started script

[rzadp]

Here are some changes to the getting-started.sh scripts we have advertised on top of the readme.

Changes to the script

1. Change echo to a more portable printf.

On my machine, the script printed a literal \n string if run with bash.
If I changed it to echo -e, then it printed a literal -e if run with sh.

Changed it to printf which is more portable.

---

2. Template selection

The script proceeded to clone and build the minimal template, which is not always what we want.
Added a selection prompt where the user can select one of the 3 templates, and choose if it should be built&run or not.
The user can also select no template at all - that way, we have a starter of a dependencies-installation script.

---

3. Added some missing dependencies for some of the systems.

A workflow testing the script

I propose a workflow, that will test the script using the expect tool.
For each OS mentioned in the script (macOS, Ubuntu, Debian, Arch, Fedora, OpenSUSE) we go through the script twice, and after that build and run the template binary.

I'm using docker containers, so we start from scratch and make sure the scripts installs all required dependencies - with the exception of macOS, which I can't run from scratch in a container.

The jobs use a selected combination of OSes, shell interpreters (bash or sh), and templates.
There is too much combinations to run them all, but I have run it once in staging to make sure all pass.

I'm adding a cron schedule because it can break without any code changes in this repository (e.g. new latest release of a container).

[kianenigma]

Template selection

In this section, when they cancel, can you also make sure to show them "if you want more templates, please see here"?

A workflow testing the script

I am not an expert to judge the complexity of the testing scripts. I can share the overall guidance that we should also not over-engineer it, both to minimize the CI costs and the maintenance of the scripts.

That is not to say the current steps are over-engineered per-se, this is more of an overall thought.

What I would suggest is only to remove the last two steps: The templates are already checked to compile, run and produce blocks elsewhere. We don't need it here again.

As noted elsewhere, if we can use the same getting started script to setup our CI environment, it will be a double win, and more incentive to battle-test it.

[kianenigma]

Some comments, but nothing major.

[rzadp]

What I would suggest is only to remove the last two steps: The templates are already checked to compile, run and produce blocks elsewhere. We don't need it here again.

Hmm, I'm not sure we're on the same page here.
Yes, the templates are checked to compile, run and produce blocks elsewhere, but only on Ubuntu/Debian and macOS.

What this introduces, is making sure that those exact lists of dependencies and Rust setup is enough to build a template on a clean arch/fedora/opensuse.

But yeah, maybe this is not a place for it. 🤔

[ggwpez]

Nice, thanks!