Core: invoke ready watcher even if discovery fails

Fixes #294

Problem

When a transport is created asynchronously via GrpcTransportBuilder.buildAsync(Runnable readyWatcher), the readyWatcher callback is never invoked if the initial discovery fails. The caller waits forever and the transport is left in an unusable state without any notification.

Root cause

YdbTransportImpl.startAsync(Runnable readyWatcher) schedules a task that calls discovery.waitReady(-1) and then readyWatcher.run():

scheduler.execute(() -> {
    discovery.waitReady(-1);
    readyWatcher.run();
});

YdbDiscovery.waitReady() throws IllegalStateException("Discovery failed", ...) when discovery fails. The exception is swallowed by the executor, so readyWatcher.run() is never executed.

Fix

Wrap waitReady in try/catch/finally so that readyWatcher.run() is always executed, and log the discovery failure with its cause for diagnostics:

scheduler.execute(() -> {
    try {
        discovery.waitReady(-1);
    } catch (RuntimeException ex) {
        logger.warn("Discovery failed during async transport initialization", ex);
    } finally {
        readyWatcher.run();
    }
});

The synchronous path (build()) is not affected: there the exception is propagated to the caller, which is the correct behavior.

Tests

Added asyncBuildDiscoveryErrorTest in YdbTransportImplTest: it stubs the discovery call to fail with UNAVAILABLE, builds the transport with buildAsync, runs the scheduler tasks and asserts that the ready watcher is invoked. The test fails without this fix (ready watcher is never called) and passes with it.

• ./mvnw -pl core test — 155 tests, all green
• Checkstyle: 0 violations