feat: Add retry option for fetching rulesets

[stephenreid]

Pull Request: Default Network Timeout and Ruleset Retry Logic

Summary

This PR introduces a default network_timeout and a configurable ruleset_id_list_retry_limit to the Ruby SDK to improve reliability during transient network instability.

1. Default Network Timeout: Set to 30 seconds. Previously, this defaulted to nil, which could lead to hung requests in certain environments depending on the underlying HTTP client's behavior.
2. Ruleset & ID List Retries: Introduced ruleset_id_list_retry_limit (defaulting to 3) to ensure that fetching configuration specs and ID lists is resilient to intermittent failures.
3. Alphabetical Organization: Reordered StatsigOptions and the corresponding documentation table alphabetically for better maintainability.

Resilience to Cloudflare 520 Errors

Cloudflare 520 ("Web Server Returned an Unknown Error") is a catch-all for unexpected responses from the origin. These are often transient and can occur during brief periods of high latency or socket hangs.

• Timeout Protection: By enforcing a 30s network_timeout, we ensure that if a Cloudflare edge or the origin hangs indefinitely (common in 520 scenarios), the SDK will proactively close the connection rather than waiting indefinitely.
• Automatic Recovery: Since 520 errors are frequently intermittent, the new retry logic allows the SDK to immediately attempt a fresh request. This significantly increases the probability of a successful initialization or sync without requiring manual intervention or app restarts.

Stability via Backoff and Jitter

The SDK utilizes an exponential backoff strategy with added jitter for these retries:

• Exponential Backoff: When a request fails, the SDK waits for an increasing amount of time before retrying (backoff * @backoff_multiplier). This prevents "thundering herd" issues where a recovering service is immediately overwhelmed by a flood of simultaneous retries from all SDK instances.
• Jitter: By incorporating randomness into the sleep interval (seen in Network#request), we ensure that multiple distributed Ruby processes don't synchronize their retry attempts. This spreads the load over time, providing the network layer and Statsig's infrastructure a better window to stabilize and process requests successfully.

Test Plan

• Verified that StatsigOptions defaults network_timeout to 30 and ruleset_id_list_retry_limit to 3.
• Added assertions to test/test_network_timeout.rb to confirm default timeout behavior.
• Added test_ruleset_id_list_retries to test/test_network.rb to verify that the SDK correctly retries failed config fetches up to the specified limit.
• Ran existing test suite to ensure no regressions in network handling.

[stephenreid]

@lfoster-statsig I explored the SDK some, the default parameter for ruleset download retries is currently 0; this can help failed initialization.

[stephenreid]

Intention implemented by d54626e#diff-af580512ac44255daab58807d7ca8efd50dce3d9fbfe7d4c96b428dd38e19d76

v2.8.4