feat: add plugin schema system, tokenizer cache, and config validation

- Add plugin schema types and runtime discovery for meta/filter plugins
- Rewrite --generate-config to use schema system instead of hardcoded types
- Add Settings::validate_config() for startup validation
- Cache tokenizer instances via static Lazy to avoid repeated BPE loading
- Add split_by_token_iter() and count_bounded() to Tokenizer
- Fix double-counting bug in TokensMetaPlugin when buffer < max_buffer_size
- Eliminate unnecessary allocations in token count methods
- Refactor token filters: remove Option<Tokenizer>, use iterator API
- Fix TailTokensFilter correctness: unbounded buffer instead of ring buffer
- Add encoding option to all token filters
- Add description() to MetaPlugin and FilterPlugin traits
- Fix unused_mut warning in compression engine (feature-gated code)

Co-Authored-By: code-review-bot <noreply@anthropic.com>

src/meta_plugin/mod.rs

@@ -479,6 +479,71 @@ where
         vec![self.meta_type().to_string()]
     }
 
+    /// Returns a description of this plugin for display in config templates.
+    ///
+    /// # Returns
+    ///
+    /// A description string (empty by default).
+    fn description(&self) -> &str {
+        ""
+    }
+
+    /// Builds the schema for this plugin from its options and outputs.
+    ///
+    /// Default implementation infers option types from YAML values and
+    /// collects enabled outputs.
+    ///
+    /// # Returns
+    ///
+    /// A `PluginSchema` describing this plugin's configuration.
+    fn schema(&self) -> crate::common::schema::PluginSchema {
+        use crate::common::schema::{OptionSchema, OptionType, OutputSchema, PluginSchema};
+
+        let options: Vec<OptionSchema> = self
+            .options()
+            .iter()
+            .map(|(key, value)| {
+                let option_type = OptionType::from_yaml_value(value);
+                let (default, required) = if value.is_null() {
+                    (None, true)
+                } else {
+                    (Some(value.clone()), false)
+                };
+                OptionSchema {
+                    name: key.clone(),
+                    option_type,
+                    default,
+                    required,
+                }
+            })
+            .collect();
+
+        let mut outputs: Vec<OutputSchema> = Vec::new();
+        for (key, value) in self.outputs() {
+            if !value.is_null() {
+                outputs.push(OutputSchema {
+                    name: key.clone(),
+                    description: key.clone(),
+                });
+            }
+        }
+        if outputs.is_empty() {
+            for output_name in self.default_outputs() {
+                outputs.push(OutputSchema {
+                    name: output_name.clone(),
+                    description: output_name,
+                });
+            }
+        }
+
+        PluginSchema {
+            name: self.meta_type().to_string(),
+            description: self.description().to_string(),
+            options,
+            outputs,
+        }
+    }
+
     /// Method to downcast to concrete type (for checking finalization state).
     ///
     /// # Returns