backend/sdoc_source_code: parse SDoc nodes from source files

[stanislaw]

This addresses #2231 for C/C++.

1. Lark parser is introduced for parsing code comments. Besides @relation markers, now the arbitrary nodes can be parsed as well, for example:

/**
 * Some text.
 *
 * @relation(REQ-1, scope=function)
 *
 * INTENTION: Intention...
 *
 * INPUT: This
 *        is
 *        a statement
 *        \n\n
 *        And this is the same statement's another paragraph.
 *
 * EXPECTED_RESULTS: This is a comment.
 */
void test_case_1(void) {
    print("hello world\n");
}

2. Using CAPITALIZED_NAME: <...> was the most straightforward both in terms of parsing and interfacing with the existing SDoc convention where the field names are capitalized. I tested \n\n to work successfully for creating paragraphs in both Doxygen and SDoc HTML.

3. Note that the strictdoc.toml has this important config option now:

source_nodes = [
  { "tests/" = { uid = "TEST_DOC", node_type = "TEST_SPEC" } }
]

...which gives a hint to the source node parser where to inject the auto-generated nodes and which type they should be (in this case description.sdoc has TEST_DOC and defines a grammar with SECTION and TEST_SPEC).

4. I was not sure how to auto-generate sections, so I went with sorting them according to the source code function paths.

The feature is super-raw but I prefer to merge it and be testing the integration early.

Related to: #2231

[stanislaw]

cc @thseiler @haxtibal @Master5