Rework a bunch of cfg(feature) flags to be more principled

[WorldSEnder]

Description

• spawn_local is now always present, so Suspension::from_future etc is not feature/platform gated anyway. The scheduler for async panics on unsupported platforms. We could go for not compiling instead when ssr without a specific platform like tokio is enabled. Having to write any(target_arch = .., feature = .. list all supported platforms) is a no-go for me. Just do what wasm_bindgen does.
• doc_test and wasm_test have been removed. The former without replacement, cfg(doctest) is stable if one needs it. The latter replaced with target_arch = "wasm32".
• trace_hydration removed. Instead, a way to get the lifecycle logs for a component was re-introduced. We should consider moving to a principled log solution, like the log crate, after measuring impact on code size instead?
• in some places, the features have been grouped in feat_ submodules and imports moved into there from top-level. Honestly lifecycle.rs is still a bit of a mess, but this PR is a start towards Feature soundness is a chore to deal with #2641

---

$ rg . -e "feature|target_arch =" --count-matches -U | awk -F: '{sum += $2} END {print sum}'
- 540
+ 499

Checklist

• I have reviewed my own code
• I have added tests

[github-actions]

Visit the preview URL for this PR (updated for commit 371a434):

https://yew-rs-api--pr2666-universal-scheduler-3dtmbqa6.web.app

_{(expires Sat, 14 May 2022 14:56:37 GMT)}

_{🔥 via Firebase Hosting GitHub Action 🌎}

[github-actions]

Size Comparison

examples	master (KB)	pull request (KB)	diff (KB)	diff (%)
boids	172.566	172.764	+0.197	+0.114%
contexts	109.438	109.633	+0.195	+0.178%
counter	86.458	86.653	+0.195	+0.226%
counter_functional	87.109	87.305	+0.195	+0.224%
dyn_create_destroy_apps	89.576	89.771	+0.195	+0.218%
file_upload	102.426	102.622	+0.196	+0.192%
function_memory_game	167.141	167.338	+0.197	+0.118%
function_router	350.237	350.424	+0.187	+0.053%
function_todomvc	161.787	161.983	+0.196	+0.121%
futures	226.421	226.617	+0.196	+0.087%
game_of_life	107.315	107.512	+0.196	+0.183%
inner_html	83.492	83.688	+0.195	+0.234%
js_callback	112.682	112.877	+0.195	+0.173%
keyed_list	194.837	195.033	+0.196	+0.101%
mount_point	86.085	86.280	+0.195	+0.227%
nested_list	115.716	115.911	+0.195	+0.169%
node_refs	90.252	90.447	+0.195	+0.216%
password_strength	1539.377	1539.575	+0.198	+0.013%
portals	96.999	97.194	+0.195	+0.201%
router	319.274	319.472	+0.197	+0.062%
simple_ssr	493.991	494.015	+0.023	+0.005%
ssr_router	425.462	425.602	+0.140	+0.033%
suspense	110.373	110.524	+0.151	+0.137%
timer	89.165	89.360	+0.195	+0.219%
todomvc	142.857	143.054	+0.196	+0.137%
two_apps	87.091	87.286	+0.195	+0.224%
web_worker_fib	153.139	153.335	+0.196	+0.128%
webgl	87.177	87.373	+0.196	+0.225%

✅ None of the examples has changed their size significantly.

[ranile]

PR looks good. Just a couple of changes to provide a better experience

We should consider moving to a principled log solution, like the log crate, after measuring impact on code size instead?

This is something I brought up before but it never got anywhere: we don't need to. The browser offers a way to filter logs in the console. We need to log at proper levels and let the browser handle the rest.

We can't provide LOG environment variable so if we end up using log, a recompile would be needed to change the log levels

[futursolo]

Could you please also update the Supported Features section in lib.rs to reflect the current status of features?

[WorldSEnder]

The failing test cases are unrelated and failing due to missing synchronization points in the ssr tests and the browser being a bit slow. Regarding the log, will maybe discussed in an upcoming issue/PR, but unrelated for now.