Replace storage-read panics with typed errors in escrow contract

[Olorunfemi20]

Closes #156

Problem

The EscrowContract read persistent storage with .unwrap() in 8 places. A
missing key (uninitialized contract, race condition, key typo, or corruption)
caused the whole transaction to panic and revert with a generic "host error",
costing users gas with no actionable reason and risking locked funds.

Changes

• Added a ContractError enum (#[contracterror], #[repr(u32)]) with 12
  descriptive, documented variants, including the required NotInitialized,
  AlreadyInitialized, Unauthorized, InvalidStatus, InvalidWinner, and
  InsufficientBalance.
• Replaced all 8 .unwrap() storage reads (and the .expect() on the pending
  operation) with typed read helpers using .ok_or(ContractError::NotInitialized).
• Every public function now returns Result<…, ContractError>; getters return
  Result<T, ContractError>.
• Converted the assert_owner/assert_contributor/assert_arbitrator/
  assert_status/assert_unlocked helpers and the timelock helpers from
  assert!/panic! to Result returns, threaded with ?.
• Validation branches now return precise errors (InvalidAmount,
  AlreadyInitialized, Unauthorized, InvalidStatus, InvalidWinner,
  PendingOperationExists, NoPendingOperation, OperationLocked,
  OperationAlreadyUnlocked, InvalidOperation).

Tests

• All #[should_panic] tests replaced with try_* calls asserting the exact
  Err variant (the token-allowance trap uses is_err()).
• Added error-path tests: fund/approve on an uninitialized contract,
  double-fund, resolve with an invalid winner, and execute with no pending
  operation.

Note: cargo was unavailable in my environment, so cargo test and
cargo clippy still need to be run locally/CI to confirm green.

[Xuccessor]

Hey 👋 Nice work on this refactor — swapping assert! panics for a typed ContractError enum and Result-returning functions is a clean improvement, and the spec updates exercise the new failure paths. Merging. 🚀