Justify our recommendation against `optional`/`provided` scope

The warning: https://github.com/jspecify/jspecify/blob/4e80178f7e3ef556169c7866e2c91dae3f69750d/docs/docs/using.md?plain=1#L32

Some previous threads offer hints of why:

• https://github.com/jspecify/jspecify/issues/389
• https://github.com/jspecify/jspecify/issues/294 (including citing Guava docs that cite how Maven doesn't have a compileOnly equivalent)

I recently learned of another, which is that kotlinc will be upset if you write fun foo() = SomeJavaThing.thatHasNullableOnItsReturnType(), as reported in https://github.com/keiyoushi/extensions-source/pull/16590:

Type annotation class 'org.jspecify.annotations.Nullable' of the inferred type is inaccessible. Check the module classpath for missing or conflicting dependencies. This will become an error in language version 2.4. See https://youtrack.jetbrains.com/issue/KT-80247.

Start with the warning in docs/docs/using.md, then read issues #389 and #294 and the linked Kotlin report for supporting context. Update the documentation so the recommendation against optional/provided scope has a clear, accurate justification, with the cited dependency and Kotlin concerns reflected; done means the guidance is understandable without relying on the issue thread.