Fix Marksman: remap heading symbol kind from String to Namespace

[EarthmanWeb]

Problem

Marksman LSP returns all markdown headings (h1–h6) with
SymbolKind.String (value 15). Serena uses symbol kind values to
classify symbols as "high-level" (classes, functions, namespaces) or
"low-level" (variables, strings, constants) for filtering purposes.
SymbolKind.String (15) falls in the low-level range, so all
markdown headings are silently filtered out of symbol overviews.

This means get_symbols_overview() on a Markdown file returns nothing,
and find_symbol cannot locate any headings by name.

Fix

Added a request_document_symbols() override in marksman.py that
post-processes the symbol tree and remaps any symbol with
SymbolKind.String → SymbolKind.Namespace (3).

Namespace is semantically appropriate for headings — they are named
sections that contain other content — and places them in the
high-level category so they appear in overviews and are findable.

The remapping is applied recursively to handle nested headings (h2
inside h1, etc.).

Example

Before:

get_symbols_overview("README.md")
# → {} (all headings filtered out — SymbolKind.String treated as low-level)

After:

get_symbols_overview("README.md")
# → {"Introduction": [...], "Installation": [...], "Usage": [...]}

[opcode81]

This sounds good. However, do you know which other types of symbols the LS assigns the String type to?
If it's only headings, then it's fine, of course.