markdown: improve fenced code block highlights

This includes an update to tree-sitter markdown that adds an `info_string`
field to `fenced_code_block`. We can use that new `info_string` field with
the negation operator to prevent highlighting fenced code blocks that have
a language string as `markup.raw.block`.

languages.toml

@@ -855,7 +855,7 @@ indent = { tab-width = 2, unit = "  " }
 
 [[grammar]]
 name = "markdown"
-source = { git = "https://github.com/MDeiml/tree-sitter-markdown", rev = "ad8c32917a16dfbb387d1da567bf0c3fb6fffde2" }
+source = { git = "https://github.com/MDeiml/tree-sitter-markdown", rev = "ab15701d8f3f68aeb74e30573b7d669a6ef2a7ed" }
 
 [[language]]
 name = "dart"

runtime/queries/markdown/highlights.scm

@@ -8,11 +8,15 @@
 (atx_heading (atx_h5_marker) @markup.heading.marker (heading_content) @markup.heading.5)
 (atx_heading (atx_h6_marker) @markup.heading.marker (heading_content) @markup.heading.6)
 
-(code_fence_content) @none
+(info_string) @label
+
+(fenced_code_block
+  !info_string
+  (code_fence_content) @markup.raw.block)
 
 [
   (indented_code_block)
-  (fenced_code_block)
+  (fenced_code_block_delimiter)
 ] @markup.raw.block
 
 (block_quote) @markup.quote

runtime/queries/markdown/injections.scm

@@ -1,5 +1,5 @@
 (fenced_code_block
-  (info_string) @injection.language
+  info_string: (info_string) @injection.language
   (code_fence_content) @injection.content
   (#set! injection.include-children))