Generalize failed-IP cooldown across all LoadBalance modes and make it configurable

[pavel-ptashyts]

The failed-IP cooldown (when a TCP connect to an IP fails, that IP is briefly
moved to the back of the failover order, then re-probed) previously lived inside
RoundRobinAddressSelector and only ran in LoadBalance.ROUND_ROBIN mode, with
a hard-coded on/off and duration.

This PR extracts the cooldown into its own mode-independent component and applies
it to any direct connection — DEFAULT mode included — so a recently-failed
IP is deprioritized on the next new connection regardless of load-balancing mode.
It also exposes the feature through client config.

What changed

• New FailedIpCooldownHolder (netty/channel) — holds the cooldown logic
  (reorder(host, addresses) + markFailed(host, address)) over bou
  The failed-IP cooldown (when a TCP connect to an IP fails, that IP is briefly
  moved to the back of the failover order, then re-probed) previously lived inside
  RoundRobinAddressSelector and only ran in LoadBalance.ROUND_ROBIN
  a hard-coded on/off and duration.

This PR extracts the cooldown into its own mode-independent component and applies
it to any direct connection — DEFAULT mode included — so a rece
IP is deprioritized on the next new connection regardless of load-balancing mode.
It also exposes the feature through client config.

What changed

• New FailedIpCooldownHolder (netty/channel) — holds the cooldown logic
  (reorder(host, addresses) + markFailed(host, address)) over bounded per-host
  state. reorder only re-orders the address list (cooling IPs to the back); it
  never drops an address, so failover always has somewhere to go.
• RoundRobinAddressSelector is reduced to pure round-robin rotation.
• New config settings (mirroring loadBalance, wired end-to-end
  AsyncHttpClientConfig, DefaultAsyncHttpClientConfig + Builder,
  AsyncHttpClientConfigDefaults, and ahc-default.properties):
  • isFailedIpCooldownEnabled() — default true
  • getFailedIpCooldownPeriod() — default PT10S
• NettyRequestSender applies the cooldown for direct connections in both
  modes (round-robin: on top of rotation, before pinning the IP-aware
  default: when ordering the resolved addresses for a new connection) and feeds
  TCP connect failures back via the connector's existing failure listener.

Behavior change

With the default failedIpCooldownEnabled=true, DEFAULT mode now
deprioritizes an IP that just failed to connect when opening its next new
connection to a multi-IP host. This preserves round-robin's previous always-on
behavior and extends the same benefit to default mode. Set
failedIpCooldownEnabled=false to restore the old default-mode behavior.

Scope & limitations

• Applies to direct connections only (no explicit Request address; no proxy,
  or proxy bypassed for the host), keyed by the target host — matchin
  round-robin eligibility. Proxied/CONNECT requests are unaffected.
• TCP-connect failures only (TLS/handshake failures are out of sc
  cooldown only re-orders — authoritative liveness is still expected at the
  DNS/resolver level. Per-host state is bounded (cap 4096) with arbitrary eviction.