docs: fix emoji rendering and simplify architecture/batch docs

Add pymdownx.emoji extension to mkdocs.yml for FontAwesome icon rendering. Simplify architecture.md diagram and remove verbose MessagePack section. Streamline batch.md round-trip explanation.

Author: howtis

docs/architecture.md

@@ -4,13 +4,13 @@ How PythonEmbed works under the hood.
 
 ## Overview
 
-PythonEmbed uses a **subprocess + MessagePack** model:
+PythonEmbed uses a **subprocess** model:
 
 ```
-┌─────────────┐     MessagePack frames      ┌──────────────────┐
-│  JVM (Java) │ ◄──────────────────────────► │ CPython process  │
-│             │     via stdin / stdout       │ (bridge.py)      │
-└─────────────┘                              └──────────────────┘
+┌─────────────┐                          ┌──────────────────┐
+│  JVM (Java) │ ◄──────────────────────► │ CPython process  │
+│             │  via stdin / stdout      │ (bridge.py)      │
+└─────────────┘                          └──────────────────┘
 ```
 
 Key design choices:
@@ -21,23 +21,7 @@ Key design choices:
 
 ## Communication Protocol
 
-All communication happens via **MessagePack** binary protocol over stdin/stdout:
-
-1. Java serializes a request (eval/exec/ref/callback/etc.) as a MessagePack frame
-2. The frame is prefixed with a 4-byte big-endian length header
-3. Python's `bridge.py` reads the length, then the full frame
-4. Python processes the request and writes a response frame back
-5. Java reads the response, deserializes, and returns the result
-
-### Why MessagePack?
-
-| Format | Size (typical) | Speed | Python support |
-|--------|---------------|-------|---------------|
-| MessagePack | ~50 bytes | Fast (binary) | `msgpack` package |
-| JSON | ~120 bytes | Medium | Built-in |
-| Pickle | ~40 bytes | Fast | Built-in (unsafe) |
-
-MessagePack provides a compact binary format with strong typing, making it well-suited for cross-process communication over stdin/stdout.
+All communication happens via a binary protocol over stdin/stdout. Each request is serialized, prefixed with a 4-byte big-endian length header, and read by `bridge.py`. Python processes the request and writes a response frame back.
 
 ## Process Lifecycle
 

docs/usage/batch.md

@@ -38,13 +38,7 @@ try (PythonEmbed py = PythonEmbed.create()) {
 
 ## Why Batch?
 
-Each `eval()`/`exec()` call involves:
-1. Serialize request (MessagePack)
-2. Write to stdin (flush)
-3. Read from stdout
-4. Deserialize response (MessagePack)
-
-Batch operations combine all requests into one MessagePack frame, eliminating N-1 round-trips.
+Each `eval()`/`exec()` call involves a full round-trip: serialize, write to stdin, read from stdout, and deserialize. Batch operations combine all requests into a single frame, eliminating N-1 round-trips.
 
 ## Batch with Pool
 

mkdocs.yml

@@ -58,3 +58,6 @@ markdown_extensions:
   - pymdownx.superfences
   - pymdownx.snippets
   - admonition
+  - pymdownx.emoji:
+      emoji_index: !!python/name:material.extensions.emoji.twemoji
+      emoji_generator: !!python/name:material.extensions.emoji.to_svg