New named parameters implementation

[nkakouros]

As already mentioned in #45, I have taken inspiration from a nice trick I found online. I expanded upon it to create the same aliases as in v2.

This is a proposal and if you find it appropriate we could work on it to expand it to fully cover v2 aliases or introduce new ones as well. For the time being I have created the [string] and [integer] aliases.

In the code, I think it would be nice to add comments to guide future contributors. I have added some minimal ones so far. Also, contrary to the previous implementation, I named internal variables that are not leaked to the calling code with friendly names to make it easier to follow the flow. Functions, aliases, etc that will be available in the scope of the caller have been namespaced with bash_oo_<library-name>. I have also set -eu (probably should have added -o pipefail). I think it would be a good idea to enable strict mode and take care of initializing our variables and handling any error. Moreover, I have used shfmt to format the source code.

I have tried to follow a TDD approach. As we have already discussed, I have used bats and a couple 3rd party libraries instead of rolling our own custom test suite. To run the tests, there is script that can be called with bin/run_tests.

[nkakouros]

I also added a travis.yml. I think I should have better reused the travis.yml from v2.

[nkakouros]

OK, I replaced my travis.yml with the one from v2.

[niieani]

Apologies that it took so long. This is a great start! I like how simple the function is. I've added one more test.

I had some trouble running bats, due to it not being compatible with macOS's readlink, but solved that by running brew install coreutils.

We should also think a bit about the target directory structure of the v3.

It would be good to have a dependency and module system in place before we go too deep with core implementation.