Fix PIBD sync stuck (#3817)

* pibd: record segment peer to exclude it from re-request

* desegmenter is blocked on a missing kernel

* segment iter

* add indexing

* pibd: log excluded peer

* segment batch optimization

* pibd: debug logging for stale requests

* pibd: filter excluded peer

* pibd: check existing kernel segment identifier

* pibd: simplify peer selection condition

---------

Co-authored-by: Joerg <wiesche89@googlemail.com>

Author: ardocrat

chain/src/pibd_params.rs

@@ -31,9 +31,12 @@ pub const KERNEL_SEGMENT_HEIGHT: u8 = 11;
 /// Maximum number of received segments to cache (across all trees) before we stop requesting others
 pub const MAX_CACHED_SEGMENTS: usize = 15;
 
+/// Number of segments to apply in a single LMDB transaction
+pub const SEGMENT_APPLY_BATCH_SIZE: usize = 4;
+
 /// How long the state sync should wait after requesting a segment from a peer before
 /// deciding the segment isn't going to arrive. The syncer will then re-request the segment
-pub const SEGMENT_REQUEST_TIMEOUT_SECS: i64 = 60;
+pub const SEGMENT_REQUEST_TIMEOUT_SECS: i64 = 20;
 
 /// Number of simultaneous requests for segments we should make. Note this is currently
 /// divisible by 3 to try and evenly spread requests amount the 3 main MMRs (Bitmap segments