Add retry mechanism for pov-recovery, fix full-node pov-recovery (#2164)

* Increase delay for pov-recovery

* Update client/service/src/lib.rs

Co-authored-by: Bastian Köcher <git@kchr.de>

* Comment

* FMT

* Clear waiting_recovery when block is recovered or recovery failed

* Introduce recovery queue that preserved insertion order

* Better error logs

* Decrease slot duration

* Style improvements

* Add option to use unordered queue

* Maintain cache of finalized blocks

* Wait for one relay chain slot before recovery

* Make retries testable

* fmt

* Improve docs

* Improve docs

* Simplify RecoveryQueue

* Remove unwanted changes

* Adjust to comments

* Apply suggestions from code review

Co-authored-by: Bastian Köcher <git@kchr.de>

* Move recovery delay into the queue

* Check for finalized number

* Clean up

* Use timer

Co-authored-by: Bastian Köcher <git@kchr.de>

* Simplify implementation

* Revert "Use timer"

This reverts commit 3809eed840d3a09d54212f99486782ff80cdc1c9.

* Properly clear `to_recover` flag

---------

Co-authored-by: Bastian Köcher <git@kchr.de>

test/service/src/cli.rs

@@ -49,6 +49,9 @@ pub struct TestCollatorCli {
 
 	#[arg(long)]
 	pub disable_block_announcements: bool,
+
+	#[arg(long)]
+	pub fail_pov_recovery: bool,
 }
 
 #[derive(Debug, clap::Subcommand)]