perf: add caching system for fast graph loading

Major performance improvements for graph visualization:

Cache System:
- Add cache.lua module that stores parsed links/tags per file
- Only re-parse files that have been modified (mtime check)
- Cache stored in .ideadrop-graph-cache.json in idea_dir
- Fast file scanning using vim.fs.find when available

Layout Optimizations:
- Reduce max iterations from 300 to 100
- Faster convergence with adjusted parameters
- Barnes-Hut approximation for large graphs (100+ nodes)
- Cache math functions locally for speed
- Skip distant node pairs in repulsion calculation
- Reuse visible_nodes array across iterations

New Commands:
- :IdeaGraph rebuild - Force full cache rebuild
- :IdeaGraphClearCache - Clear cache file

This makes opening the graph nearly instant for previously
scanned vaults, similar to Obsidian's behavior.

lua/ideaDrop/utils/constants.lua

@@ -146,18 +146,21 @@ M.DEFAULT_KEYMAPS = {
 
 -- Graph visualization settings
 M.GRAPH_SETTINGS = {
-	-- Layout algorithm parameters
+	-- Layout algorithm parameters (optimized for speed)
 	LAYOUT = {
 		-- Fruchterman-Reingold parameters
-		REPULSION_STRENGTH = 5000, -- How strongly nodes repel each other
-		ATTRACTION_STRENGTH = 0.01, -- Spring constant for connected nodes
-		IDEAL_EDGE_LENGTH = 50, -- Ideal distance between connected nodes
-		GRAVITY = 0.1, -- Pull toward center
-		DAMPING = 0.85, -- Velocity damping per iteration
-		MIN_VELOCITY = 0.01, -- Stop threshold
-		MAX_ITERATIONS = 300, -- Maximum layout iterations
-		COOLING_RATE = 0.95, -- Temperature cooling per iteration
-		INITIAL_TEMPERATURE = 100, -- Initial movement freedom
+		REPULSION_STRENGTH = 3000, -- How strongly nodes repel each other
+		ATTRACTION_STRENGTH = 0.02, -- Spring constant for connected nodes
+		IDEAL_EDGE_LENGTH = 40, -- Ideal distance between connected nodes
+		GRAVITY = 0.15, -- Pull toward center (stronger = faster convergence)
+		DAMPING = 0.8, -- Velocity damping per iteration
+		MIN_VELOCITY = 0.5, -- Stop threshold (higher = faster stop)
+		MAX_ITERATIONS = 100, -- Maximum layout iterations (reduced for speed)
+		COOLING_RATE = 0.9, -- Temperature cooling per iteration (faster cooling)
+		INITIAL_TEMPERATURE = 80, -- Initial movement freedom
+		-- Barnes-Hut optimization threshold (for large graphs)
+		BARNES_HUT_THETA = 0.8, -- Use approximation for distant nodes
+		LARGE_GRAPH_THRESHOLD = 100, -- Use optimizations above this node count
 	},
 
 	-- Visual settings