feat: add retry with exponential backoff to Soroban RPC client

[rilwanubala]

Summary

Closes #199

Adds retry with exponential backoff to all Soroban RPC calls in backend/src/soroban.rs. Transient failures (timeouts, connection resets, HTTP 5xx) are automatically retried up to 3 times with doubling delays capped at 30 seconds total. Non-transient errors (4xx, contract-level NOT_FOUND / UNAUTHORIZED / NOT_PAID) are surfaced immediately without retrying.

What changed

Core implementation (backend/src/soroban.rs)

• is_transient(err) — classifies errors as retryable vs. permanent
• with_retry(op) — generic async retry helper (max 3 retries, 200 ms → 400 ms → 800 ms backoff, 30 s total cap)
• rpc_post(req) — internal HTTP helper wrapping every RPC call with retry; treats HTTP 5xx as transient
• check_horizon_health() — also wrapped with retry for 5xx responses

Bug fixes (pre-existing compile/test failures)

• health.rs: Fix incompatible if/else return types in mock server + missing into_make_service() in axum::serve
• cancel.rs, refund.rs: SorobanClient::new called with 2 args instead of 3
• pay/cancel/refund tests: Accept 415 or 422 for missing-body requests (axum-test v14 sends no Content-Type)
• routes/mod.rs: Add missing cancel module export
• main.rs: Add missing cancel_invoice / refund_invoice imports
• Rename routes.rs → routes_auth_legacy.rs to remove Rust E0761 ambiguity with routes/mod.rs

Tests

All 20 tests pass (cargo test):

running 20 tests
test soroban::tests::transient_false_for_not_found ... ok
test soroban::tests::transient_false_for_unauthorized ... ok
test soroban::tests::transient_false_for_not_paid ... ok
test soroban::tests::transient_false_for_empty_rpc_result ... ok
test soroban::tests::transient_true_for_5xx_message ... ok
test soroban::tests::transient_true_for_timeout_message ... ok
test soroban::tests::transient_true_for_connection_message ... ok
test soroban::tests::retries_on_5xx_and_eventually_succeeds ... ok
test soroban::tests::gives_up_after_max_retries_on_persistent_5xx ... ok
test soroban::tests::does_not_retry_on_4xx ... ok
test soroban::tests::get_invoice_retries_on_5xx_and_succeeds ... ok
test soroban::tests::not_found_error_is_not_retried ... ok
test routes::cancel::tests::test_cancel_invoice_missing_body_returns_422 ... ok
test routes::cancel::tests::test_cancel_invoice_unreachable_rpc_returns_error ... ok
test routes::health::tests::returns_200_when_all_dependencies_are_healthy ... ok
test routes::health::tests::returns_503_when_any_dependency_is_degraded ... ok
test routes::pay::tests::test_pay_invoice_missing_body_returns_422 ... ok
test routes::pay::tests::test_pay_invoice_unreachable_rpc_returns_5xx_or_404 ... ok
test routes::refund::tests::test_refund_invoice_missing_body_returns_422 ... ok
test routes::refund::tests::test_refund_invoice_unreachable_rpc_returns_error ... ok

test result: ok. 20 passed; 0 failed; 0 ignored; 0 measured

New tests added (soroban::tests)

Test	What it verifies
transient_false_for_not_found	NOT_FOUND is not retried
transient_false_for_unauthorized	UNAUTHORIZED is not retried
transient_false_for_not_paid	NOT_PAID is not retried
transient_false_for_empty_rpc_result	Empty result is not retried
transient_true_for_5xx_message	5xx message is classified transient
transient_true_for_timeout_message	timeout message is classified transient
transient_true_for_connection_message	connection error is classified transient
retries_on_5xx_and_eventually_succeeds	Mock server: 2×503 then success → 3 total requests
gives_up_after_max_retries_on_persistent_5xx	Mock server: all 503s → exactly MAX_RETRIES+1 requests
does_not_retry_on_4xx	Mock server: 404 → exactly 1 request (no retry)
get_invoice_retries_on_5xx_and_succeeds	get_invoice retries correctly
not_found_error_is_not_retried	Contract NOT_FOUND → single attempt only

Retry behaviour

Condition	Behaviour
HTTP 5xx from RPC node	Retry (up to 3×, backoff 200→400→800 ms)
Connection refused / timeout	Retry
HTTP 4xx from RPC node	No retry
Contract NOT_FOUND / UNAUTHORIZED / NOT_PAID	No retry
Empty RPC result	No retry
Total retry budget exceeded (>30 s)	Stop retrying, surface error

[drips-wave]

@rilwanubala Great news! 🎉 Based on an automated assessment of this PR, the linked Wave issue(s) no longer count against your application limits.

You can now already apply to more issues while waiting for a review of this PR. Keep up the great work! 🚀

Learn more about application limits