feat: allows to propagete providers named reset to its dependents

[esroyo]

This PR implements the ideas proposed on issue #193

@young-steveo I think the most "notable" change, is that It is no longer possible to configure the final instance as a direct value in the property descriptor. Regardless of the existence of middleware, the getter is needed to keep track of the accesses to the container.

Although that is an "internal detail", that fact could be relevant enough to represent a major version bump.

Changes to the README are fairly minimal; awaiting for your feedback before investing more work. Thanks!

[young-steveo]

This looks good! I added a couple of comments. Also, could you elaborate on

It is no longer possible to configure the final instance as a direct value in the property descriptor

I'm not sure I follow.

[esroyo]

It is no longer possible to configure the final instance as a direct value in the property descriptor

I'm not sure I follow.

I was alluding to the elimination of the possibility to configure value in the property descriptor here: https://github.com/young-steveo/bottlejs/pull/196/files#diff-fdb247437fd68a61581f78513f2d7b710cfc757b6288e57cfee6e9182d8678fcL30
We would need to go always via the getter function, which in the end is an extra function call for cases when there is no middleware.

[young-steveo]

I'm going to call it good. I'm unsure if this requires a major version bump, but I'll investigate before figuring out which version to deploy.

Thanks for all your work!

[esroyo]

I'm going to call it good. I'm unsure if this requires a major version bump, but I'll investigate before figuring out which version to deploy.

Thanks for all your work!

My two cents about the version bump:

• technically this is a clean API addition and would fit in a minor bump.
• changing a function from one param to two in JS is always delicate. There is a risk that someone has been passing around the function in callback style to Array prototype methods, which usually work nice with one param functions but can introduce mysterious behaviours when using more than one param (typical Array#map(parseInt) problem). Probably not the case for resetProviders, but you never know.
• Given the chances of this package being used on relatively old or stable code bases, I would opt for a major bump to ensure no risk of unintentional upgrades on those code bases.

I'm grateful to have the opportunity to contribute to this little piece of software, which I have used so much. Thanks to you @young-steveo for the original creation and for keeping this package alive during all this years.

Best regards.