Refactor LSP manager architecture and event + completion pipeline

- Replace legacy LSPManagerClient + LSPManagerUI with ClientManager and UIManager
- Remove WebsocketClient in favor of unified Websocket implementation
- Move LSP initialization config loading into centralized config module
- Update LSPClient to support dynamic socket, improved init params, and doc version tracking
- Introduce range-based didChange notifications and add implementation + references requests
- Expand LSPClientEvents to support implementation/references and structured range edits
- Simplify websocket response handling and normalize LSP response parsing flow
- Decouple UI from LSP manager core; UI now emits address/port for client creation
- Refactor completion provider pipeline:
  - Split TextChangedEvent into TextInsertedEvent and DeleteRangeEvent
  - Update ProviderResponseCacheBase and controller dispatch paths accordingly
- Improve SourceBuffer and SourceFile event tracking with delete-range support
- Update plugin system:
  - Centralize command/provider registration via helper methods
  - Add LSP commands: definition, references, implementation, toggle UI
- Enhance response handlers to support references and implementation hooks
- Improve Python LSP config (jedi completion, signatures, references enabled)
- Fix minor GTK lifecycle and buffer signal handling issues
- Clean up unused imports, dead code, and outdated JSON server configs

plugins/code/language_server_clients/lsp_manager/provider/provider.py

@@ -32,20 +32,7 @@ class Provider(GObject.GObject, GtkSource.CompletionProvider):
         return "LSP Code Completion"
 
     def do_match(self, context):
-        # Note: If provider is in interactive activation then need to check
-        # view focus as otherwise non focus views start trying to grab it.
-        # completion = context.get_property("completion")
-        # if not completion.get_view().has_focus(): return
-
         iter = self.response_cache.get_iter_correctly(context)
-        iter.backward_char()
-        ch = iter.get_char()
-
-        # NOTE: Look to re-add or apply supporting logic to use spaces
-        # As is it slows down the editor in certain contexts...
-        # if not (ch in ('_', '.', ' ') or ch.isalnum()):
-        if not (ch in ('_', '.') or ch.isalnum()):
-            return False
 
         buffer = iter.get_buffer()
         if buffer.get_context_classes_at_iter(iter) != ['no-spell-check']:
@@ -68,6 +55,7 @@ class Provider(GObject.GObject, GtkSource.CompletionProvider):
 #        return GtkSource.CompletionActivation.NONE
         return GtkSource.CompletionActivation.USER_REQUESTED
 #        return GtkSource.CompletionActivation.INTERACTIVE
+#        return GtkSource.CompletionActivation.USER_REQUESTED | GtkSource.CompletionActivation.INTERACTIVE
 
     def do_populate(self, context):
         results   = self.response_cache.filter_with_context(context)

plugins/code/language_server_clients/lsp_manager/provider/provider_response_cache.py

@@ -33,9 +33,13 @@ class ProviderResponseCache(ProviderResponseCacheBase):
         if self.lsp_manager_client:
             self.lsp_manager_client.process_file_save(event)
 
-    def process_file_change(self, event):
+    def process_file_text_inserted(self, event):
         if self.lsp_manager_client:
-            self.lsp_manager_client.process_file_change(event)
+            self.lsp_manager_client.process_file_text_inserted(event)
+
+    def process_file_delete_range(self, event):
+        if self.lsp_manager_client:
+            self.lsp_manager_client.process_file_delete_range(event)
 
     def filter(self, word: str) -> list[dict]:
         return []