Register Jackson modules, based on global user configuration

[nipafx]

One way to address #624 (JSON argument source fails with LocalDate property deserialization) is for Pioneer to globally register Jackson modules based on a configuration option set by the user. The option is necessary because there are different possible behaviors (e.g. only register well-known modules vs registering all modules) that don't work equally well across all test suites.

After conversations in #624 and #629, I propose the following:

• add configuration parameter junitpioneer.jackson.modules.registration
• implement strategies for the values none, well_known, all

The current behavior is none and since switching to any other would be a backwards-incompatible change, I recommend to have none as default for now. Overall, I think well_known is the better default, though, so I propose we make that the default in 3.0 and already document that now.

---

If this proposal is accepted and implemented, a follow-up issue could add a more sophisticated option:

• for junitpioneer.jackson.modules.registration, add the new value list
• add new parameter junitpioneer.jackson.modules.list
• implement a strategy that if ...modules.registration is set to list, the value of ...modules.list is interpreted as a comma-separated list of Jackson modules that will be registered
• if ...modules.list is set and ...modules.registration is set to any other value than list, there should be a warning or even error

I only added this additional proposal here to foreshadow a more detailed option in case the three values none, well_known, all aren't flexible enough, so for this issue we can focus on implementing those three.

[nipafx]

(Not up for grabs yet. Need to discuss it first.)

[meredrica]

Another approach would be to simply allow to list the classes that should be registered. Grab the list, instanciate via reflection, crash on wrong input. This would be future proof and allows to register custom modules etc without having to go down the route of defining what well known means.

[nipafx]

That's what I meant with the follow-up proposal (under the vertical divider).

[Bukama]

I agree with @nipafx. The defensive approach with none as the default should be the way to go.

[meredrica]

That's what I meant with the follow-up proposal (under the vertical divider).

I'm sorry, I have absolutely no idea why I didn't see this.

I agree that none should be the default.

[cwensel]

I like the idea of having the list escape hatch.