Unit tests that read prefs can randomly panic.

[emilio]

See for example the failure I got in #20708:

---- viewport::cascading_within_viewport_rule stdout ----
	thread 'viewport::cascading_within_viewport_rule' panicked at 'Resource reader not set.', libcore/option.rs:914:5
note: Some details are omitted, run with `RUST_BACKTRACE=full` for a verbose backtrace.
stack backtrace:
   0: std::sys::unix::backtrace::tracing::imp::unwind_backtrace
             at libstd/sys/unix/backtrace/tracing/gcc_s.rs:49
   1: std::sys_common::backtrace::print
             at libstd/sys_common/backtrace.rs:71
             at libstd/sys_common/backtrace.rs:59
   2: std::panicking::default_hook::{{closure}}
             at libstd/panicking.rs:206
   3: std::panicking::default_hook
             at libstd/panicking.rs:216
   4: std::panicking::rust_panic_with_hook
             at libstd/panicking.rs:400
   5: std::panicking::begin_panic_fmt
             at libstd/panicking.rs:347
   6: rust_begin_unwind
             at libstd/panicking.rs:323
   7: core::panicking::panic_fmt
             at libcore/panicking.rs:71
   8: core::option::expect_failed
             at libcore/option.rs:914
   9: <core::option::Option<T>>::expect
             at /checkout/src/libcore/option.rs:302
  10: embedder_traits::resources::read_bytes
             at components/embedder_traits/resources.rs:20
  11: embedder_traits::resources::read_string
             at components/embedder_traits/resources.rs:24
  12: core::ops::function::FnOnce::call_once
             at components/config/prefs.rs:21
             at /checkout/src/libcore/ops/function.rs:223
  13: <lazy_static::lazy::Lazy<T>>::get::{{closure}}
             at /home/servo/.cargo/registry/src/github.com-1ecc6299db9ec823/lazy_static-1.0.0/src/lazy.rs:24
  14: std::sync::once::Once::call_once::{{closure}}
             at /checkout/src/libstd/sync/once.rs:227
  15: std::sync::once::Once::call_inner
             at libstd/sync/once.rs:340
  16: std::sync::once::Once::call_once
             at /checkout/src/libstd/sync/once.rs:227
  17: <servo_config::prefs::PREFS as core::ops::deref::Deref>::deref
             at /home/servo/.cargo/registry/src/github.com-1ecc6299db9ec823/lazy_static-1.0.0/src/lazy.rs:23
             at /home/servo/buildbot/slave/linux-dev/build/<__lazy_static_internal macros>:13
             at /home/servo/buildbot/slave/linux-dev/build/<__lazy_static_internal macros>:14
  18: style_tests::viewport::test_viewport_rule
             at tests/unit/style/viewport.rs:48
  19: style_tests::viewport::cascading_within_viewport_rule
             at tests/unit/style/viewport.rs:198
  20: style_tests::__test::TESTS::{{closure}}
             at tests/unit/style/viewport.rs:194
  21: core::ops::function::FnOnce::call_once
             at /checkout/src/libcore/ops/function.rs:223
  22: <F as alloc::boxed::FnBox<A>>::call_box
             at libtest/lib.rs:1452
             at /checkout/src/libcore/ops/function.rs:223
             at /checkout/src/liballoc/boxed.rs:635
  23: __rust_maybe_catch_panic
             at libpanic_unwind/lib.rs:101

This test reads a pref to see if the viewport rule is enabled, but #20533 made it so that only calls to PREFS.set were guarder by register_resources_for_tests.

register_resources_for_tests is not great: any patch that adds something that reads a pref on a code path which is unit-tested may make tests panic randomly.

@paulrouget, this is a regression from #20533, mind looking into it? I think we could have a "test" feature for the resources dir that did what register_resources_for_tests lazily... Or something.

cc @mbrubeck