reduce edition requirement for keyword test

[tshepang]

Also, add Reference ids

[rustbot]

r? @petrochenkov

rustbot has assigned @petrochenkov.
They will have a look at your PR within the next two weeks and either review your PR or reassign to another reviewer.

Use r? to explicitly pick a reviewer

[petrochenkov]

• I don't see any existing uses of //@ edition ranges in the test suite.
  When is it supposed to be used? Why in this test specifically?
• When are the //@ reference directives supposed to be used?
  If ~all tests will have them then it's fine, otherwise it's not clear why this test should have them. It's not directly related to some reference clause, it's not even useful in general, and is mostly an example of reddit-oriented testing.

[tshepang]

• 2018.. means the test is applicable to 2018 edition onwards
  it is a recent feature, added in Implement range support in //@ edition #146166

• the test will be linked on these clauses

  • https://doc.rust-lang.org/nightly/reference/keywords.html#r-lex.keywords.strict
  • https://doc.rust-lang.org/nightly/reference/keywords.html#r-lex.keywords.weak

  the idea is to eventually have Reference have tests associated with a bunch of its clauses

[petrochenkov]

I've seen #146166, but it doesn't answer the question.
I'd expect the edition ranges to be (semi) automatically mass-annotated as a part of #145364.

For addition of the reference annotations I'd also expect to be done more systematically, similarly to #132376, but that's probably less likely to happen.

[petrochenkov]

The change itself is harmless, but it's sort of like manually formatting code one function at a time instead of rustfmting the whole codebase, I don't want to encourage this.

[tshepang]

• lowered edition to 2018 because that test can run on that edition
• made it a range because the test can also run on later editions
• I do not expect an automatic edition mass-annotation to lower the edition of this test automatically
• I added a reference id since I was touching it anyway... doing it more systematically feels much out of scope