fix: Terminal escape sequence responses displayed on first prompt when starting tmux (#90)

* fix: Replace crossterm event parsing with raw stdin reading to prevent terminal escape sequence corruption

When running tmux inside bssh PTY sessions, terminal escape sequence responses
(DA1, DA2, DA3, XTGETTCAP) were appearing as raw text on the first prompt.
This occurred because crossterm::event::read() parses escape sequences,
consuming the ESC byte (0x1b) and corrupting terminal responses.

Solution:
- Keep using crossterm for terminal mode management (enable_raw_mode/disable_raw_mode)
- Replace crossterm::event::read() with direct stdin.read() using nix::poll for polling
- Add RawInputReader module for transparent byte passthrough
- Add LocalEscapeDetector for OpenSSH-style ~. disconnect handling

Changes:
- Add nix crate dependency with poll feature
- Create src/pty/session/raw_input.rs: Raw byte input reader
- Create src/pty/session/local_escape.rs: Local escape sequence detector
- Modify session_manager.rs: Use RawInputReader instead of crossterm event parsing
- Update mod.rs: Include new modules

This approach matches OpenSSH's behavior, providing transparent passthrough
of all bytes including terminal responses, arrow keys, function keys, and
other escape sequences.

Fixes #87

* docs: Add PTY escape sequence documentation for ~. disconnect

* fix: Address code review feedback - improve safety docs and suppress dead code warnings

* update: cargo fmt

* fix: Correct doc comment overindentation in raw_input.rs

Fix clippy doc_overindented_list_items warning by reducing
indentation of continued list item text from 14 to 2 spaces.

* feat: Add comprehensive unit tests for PTY session modules

Add missing test coverage for PTY-related modules:

- RawInputReader: 10 new tests for poll timeout clamping,
  boundary conditions (u16::MAX), and duration handling
- TerminalModes: 18 new tests verifying control characters,
  signal generation, canonical mode, echo, baud rates,
  flow control, and mode completeness
- LocalEscapeDetector: 19 new edge case tests for buffer
  boundaries, CR/LF handling, Unicode, and state transitions
- EscapeSequenceFilter: 32 new tests for buffer boundaries,
  consecutive sequences, OSC/DCS filtering, and stress scenarios

Total: 79 new tests, 858 lines of test code added.

Author: inureyes

Cargo.lock

@@ -43,6 +43,7 @@ regex = "1.12.2"
 lazy_static = "1.5"
 ctrlc = "3.5.1"
 signal-hook = "0.3.18"
+nix = { version = "0.29", features = ["poll"] }
 atty = "0.2.14"
 arrayvec = "0.7.6"
 smallvec = "1.15.1"

Cargo.toml

@@ -101,6 +101,14 @@ bssh -o StrictHostKeyChecking=no user@host
 bssh -Q cipher
 ```
 
+**PTY Session Escape Sequences:**
+
+Like OpenSSH, bssh supports escape sequences in PTY sessions. These must be typed at the beginning of a line (after pressing Enter):
+
+| Escape | Description |
+|--------|-------------|
+| `~.`   | Disconnect from the remote host |
+
 ### Port Forwarding
 ```bash
 # Local port forwarding (-L)

README.md

@@ -13,28 +13,32 @@
 // limitations under the License.
 
 //! Terminal constants and key sequence definitions
+//!
+//! NOTE: Many key sequence constants are currently unused since we switched to
+//! raw byte passthrough (see issue #87), but are kept for reference and potential
+//! future debugging use.
+
+// Allow dead code for unused key sequence constants
+#![allow(dead_code)]
 
 // Buffer size constants for allocation optimization
 // These values are chosen based on empirical testing and SSH protocol characteristics
 
 /// Maximum size for terminal key sequences (ANSI escape sequences are typically 3-7 bytes)
 /// Value: 8 bytes - Accommodates the longest standard ANSI sequences (F-keys: ESC[2x~)
 /// Rationale: Most key sequences are 1-5 bytes, 8 provides safe headroom without waste
-#[allow(dead_code)]
 pub const MAX_KEY_SEQUENCE_SIZE: usize = 8;
 
 /// Buffer size for SSH I/O operations (4KB aligns with typical SSH packet sizes)
 /// Value: 4096 bytes - Matches common SSH packet fragmentation boundaries
 /// Rationale: SSH protocol commonly uses 4KB packets; larger buffers reduce syscalls
 /// but increase memory usage. 4KB provides optimal balance for interactive sessions.
-#[allow(dead_code)]
 pub const SSH_IO_BUFFER_SIZE: usize = 4096;
 
 /// Maximum size for terminal output chunks processed at once
 /// Value: 1024 bytes - Balance between responsiveness and efficiency
 /// Rationale: Smaller chunks improve perceived responsiveness for interactive use,
 /// while still being large enough to batch terminal escape sequences efficiently.
-#[allow(dead_code)]
 pub const TERMINAL_OUTPUT_CHUNK_SIZE: usize = 1024;
 
 /// PTY message channel sizing:
@@ -57,7 +61,7 @@ pub const INPUT_POLL_TIMEOUT_MS: u64 = 500;
 /// - Tasks should check cancellation signal frequently (10-50ms intervals)
 pub const TASK_CLEANUP_TIMEOUT_MS: u64 = 100;
 
-// Const arrays for frequently used key sequences to avoid repeated allocations
+// Const arrays for frequently used key sequences to avoid repeated allocations.
 /// Control key sequences - frequently used in terminal input
 pub const CTRL_C_SEQUENCE: &[u8] = &[0x03]; // Ctrl+C (SIGINT)
 pub const CTRL_D_SEQUENCE: &[u8] = &[0x04]; // Ctrl+D (EOF)