Document how lints are actually emitted

[rylev]

The section on linting describes how to declare a lint and how lint passes are constructed, but we are still missing an explanation on how the lint system takes buffered lints and decides whether to actually emit them to users. For instance, how does --cap-lints work?

[jieyouxu]

Jotting down some notes for involved concepts:

• Lint passes (EarlyLintPass vs LateLintPass)
• Lint groups
• Lint levels
• In-source annotations versus cli flags
• Lint level precedence (e.g. multiple annotations on the same node, scoping)
• CLI -A/W/D/F <lint_name> flags
• --cap-lints
• --force-warn
• warnings special speudo lint group
• Lints tied to editions
• Builtin lints
• Pre-expansion lints (deprecated but still a thing)
• Lint buffers (e.g. in rustc_resolve)
• Diagnostics infra's interaction with lints (mapping from lint levels to diag levels, emitters)

[jieyouxu]

(BTW this is very underspecified and underdocumented, see #113307 where I found out rust-lang/rust#113307 (comment))