UCT/IB/RC: add adaptive TX CQ moderation

When many RC endpoints each send fewer messages than the
tx_moderation threshold, no signaled sends are generated and TX CQ
credits are never reclaimed, eventually exhausting the TX buffer pool.

Fix by also checking iface-level cq_available in the moderation
decision. When CQ credits drop to the tx_moderation threshold, force
the next send on any endpoint to be signaled, ensuring timely
completion processing regardless of per-endpoint send counts.

Closes #1307

Author: Nathan

src/uct/ib/rc/base/rc_ep.h

@@ -469,7 +469,9 @@ uct_rc_txqp_completion_inl_resp(uct_rc_txqp_t *txqp, const void *resp, uint16_t
 static UCS_F_ALWAYS_INLINE uint8_t
 uct_rc_iface_tx_moderation(uct_rc_iface_t *iface, uct_rc_txqp_t *txqp, uint8_t flag)
 {
-    return (txqp->unsignaled >= iface->config.tx_moderation) ? flag : 0;
+    return (txqp->unsignaled >= iface->config.tx_moderation) ||
+           (iface->tx.cq_available <= (signed)iface->config.tx_moderation)
+           ? flag : 0;
 }
 
 static UCS_F_ALWAYS_INLINE void