Improve presentation of file system mapping

[henrikt-ma]

This is a follow-up to #3885 with general cleanup of the the presentation in File System Mapping of Package/Class.

Improvements include:

• More careful introduction and use of terminology (structured entity and nonstructured entity).
• Aligned language in subsections to make similarities easier to spot.
• Add missing rule that package.mo should only define a single class.
• Expand non-normative section on sub-directories without package.mo.

[HansOlsson]

I believe we need to discuss this more - as the PR made me realize that it is currently a bit of a mess.

So, the issue is that a .mo file can be:

• A file (not called package.mo) containing a top-level package or class
• A package.mo file for the class corresponding to the directory name
• A .mo file inside a directory with a package.mo

To me both of the latter ones are part of "structured entity"; i.e, the entire directory from the top-down would be called a structured entity. (Note that we have package.mo both at the top and further down; with similar rules - although only the top-directory may include version.)

[HansOlsson]

Language group: should ideally to re-work to avoid the terms structured and non-structured entity.

[HansOlsson]

As noted I would only call single files "nonstructured entity", so not call these files "nonstructured entity".

[HansOlsson]

I tried to look hard at the current specification and it seems to deliberately avoid using the term "nonstructured entity" for files that are part of directories with package.mo:

Specifically,

• A directory is called "structured entity"
• A single file is a "nonstructured entity"
• The other cases are called "sub-entity of an enclosing structured entity"

[HansOlsson]

The nonstructured entity is a bit of a mess, and this PR doesn't resolve that.
To me it should be reserved for single-file-mapping, and we don't need to define structured and nonstructured entity.