Fix doc attribute parsing to properly handle block comments

[sd-collins]

The existing parsing logic is flawed, as it concatenates consecutive lines, e.g. this:

/// # Heading
/// This is the first paragraph.
///
/// - First
/// - Second
/// - Third

...becomes this (correct):

/**
 * # Heading
 * This is the first paragraph.
 *
 * - First
 * - Second
 * - Third
 */

But the same thing as a block comment:

/**
# Heading
This is the first paragraph.

- First
- Second
- Third
*/

...becomes this (broken):

/**
 * # Heading This is the first paragraph.
 *
 *- First - Second - Third
 */

The current splitting logic (introduced in #454) tries to join together lines in the same paragraph but without properly parsing and interpreting the markdown, leading to removing semantically-important line breaks.
It also allows for broken syntax if a / is placed at the beginning of a line, since there is no leading space after the * marker (creating a */ which closes the comment block).

The simpler approach in this PR is to simply split on newlines, which makes the parsing behaviour of block doc comments more consistent with single-line doc comments. We also add a single space at the start of each line, which is normally included in single-line doc comments but not in the block variant and prevents us accidentally creating a */.

EDIT: I also realised I had to remove "common" whitespace between all lines in the block, or else block doc comments within modules are parsed with leading whitespace included, which is usually not intended. I have included tests for all new functionality and slightly modified the documentation_attr test for the new behaviour.

[emilio]

Looks sensible