ADK in general and FunctionTool must allow injecting one's custom ObjectMapper instead of hard-coding it

[vorburger]

Describe the bug

ADK currently hard-codes new ObjectMapper().registerModule(new Jdk8Module()) in places like here.

This is well intended ("it works out of the box"), but it's not extensible, and blocks one from using a custom ObjectMapper; e.g. in my project I would like (have) to use this one instead; notably for:

mapper.registerModule(new JavaTimeModule());

But this (the JavaTimeModule) is really just one case - the point is that users should be able to inject their own.

The specific issue I've faced is that my ToDoTool here used a domain class ToDo with Instant fields, and that made choke.

I'm temporarily working around it with enola-dev/enola#1855 - but I shouldn't have to.

To Reproduce
Steps to reproduce the behavior:

1. git clone https://github.com/enola-dev/enola.git
2. Revert fix: Work around https://github.com/enola-dev/enola/issues/1854 enola-dev/enola#1855
3. Run b test //java/dev/enola/todo/...
4. See error

Expected behavior
A clear and concise description of what you expected to happen.

Additional context

@Poggecci @Doris26 @shukladivyansh @Jacksunwei @glaforge agreed that this makes sense?

[vorburger]

I've just remember that there is, of course, the JsonBaseModel already; so I think that this issue is actually 2 somewhat separate things:

a) ADK should consistently use that everywhere (IMHO) - agreed? It currently does not e.g. here, and possibly/likely elsewhere (I've not had a closer look)

b) ADK should then, additionally, also, let users inject (pass through) their own ObjectMapper in relevant methods / classes

Does this make sense?