Transition from Travis -> GH Actions and use Ameba

[stephendolan]

This PR tackles the following:

• Move from Travis CI to GitHub Actions. This is quickly becoming the most popular way to do CI in the GitHub ecosystem, and there are tons of built-in tools that will only get better with time. I've set this up identically to the Travis file, where we test against both the latest and nightly Crystal versions, allowing nightly to fail as an "experimental" version.
• Add Dependabot capabilities to submit PRs for GitHub action library updates
• Add badges to the README to show the current CI status and latest GitHub release version
• Add Ameba linting to our specs. This helps to avoid bikeshedding on code style in favor of a popular community set of standards.
• Removed crystal-coverage shard from development dependencies since it didn't appear to have been used previously with Travis, and doesn't currently compile.

[stephendolan]

@confact, here's an output sample from Ameba to get a feel for what I'd be changing if we add it to the repo:

~/Repos/@oss/stripe.cr(sd-github-actions-switch) » ./bin/ameba                                                           stephen@stephens-mbp
Inspecting 94 files.

......................FFFFFFFFFFFFFFFFFFFFFFFFF.FFFFFFFFFFFFFFFFFFFF..........................

src/stripe/methods/core/customers/tax_ids/create_customer_tax_id.cr:19:7
[C] Style/RedundantReturn: Redundant `return` detected
> return Customer::TaxID.from_json(response.body)
  ^
src/stripe/methods/core/customers/tax_ids/retrieve_customer_tax_id.cr:9:7
[C] Style/RedundantReturn: Redundant `return` detected
> return Customer::TaxID.from_json(response.body)
  ^
src/stripe/methods/core/customers/update_customer.cr:18:5
[W] Lint/UselessAssign: Useless assignment to variable `default_source`
> default_source = default_source.as(Token).id if default_source.is_a?(Token)
  ^

Finished in 149.18 milliseconds 

94 inspected, 87 failures.

[confact]

@stephendolan, that's all fine. I know Ameba is starting to become standard in crystal, if not already. So go for it. Just happy that you want to spend time on this :)

[stephendolan]

That sounds great, @confact. One other question since I'm far from a semantic versioning expert...

Do you think this kind of refactor to satisfy Ameba is worth a version bump, or as long as all of the tests pass should we just leave the version alone?

[stephendolan]

Nevermind, I just realized that there's literally no point in cutting a release just for this CI and Ameba stuff :)

I did end up opening crystal-ameba/ameba#194 since I think there's an issue with Ameba that's preventing this from passing. Once that issue is resolved, I'll submit a follow up PR to remove the .ameba.yml.

[stephendolan]

The nightly build will continue to fail until we figure out how to refactor Stripe::Error not to use JSON::Mapping and instead use JSON::Serializable. That said, I've marked nightly as experimental, so we should probably just merge this one and tackle that refactor in a separate PR.

This one just looks huge because I finally got crystal docs working, which allowed me to push an update to all of those committed doc files. Once we get automatic documentation working, they'll all be removed.

[confact]

@stephendolan this looks good to me, We can find some solution for the error later. It would be nice for everything to pass but let's also keep this small enough :)

Great job with docs as well, just prove it is good with some new eyes and energy to look things through!

[confact]

LGTM