fix: ensure that downloader sync gap is only set once

[klkvr]

This is another followup around #14288

Here we are polling not the ReverseHeaderDownloader but actually spawned downloader:

reth/crates/stages/stages/src/stages/headers.rs

Lines 234 to 239 in ff404c8

	// let the downloader know what to sync
	self.downloader.update_sync_gap(gap.local_head, gap.target);
	// We only want to stop once we have all the headers on ETL filespace (disk).
	loop {
	match ready!(self.downloader.poll_next_unpin(cx)) {

which means that if we're about to poll a DetachedHead error, we will always end up calling update_sync_gap right before it, thus making downloader starting to sync again.

This would result in next stage run getting the same error again, and causing some weird issues like that one:

2025-06-05T19:20:15.570995Z  INFO node{idx=0}: reth_node_events::node: Preparing stage pipeline_stages=1/12 stage=Headers checkpoint=28 target=None
2025-06-05T19:20:15.571018Z DEBUG node{idx=0}:ChainOrchestrator::poll: sync::stages::headers: Commencing sync tip=Hash(0x50370fd9a6b5bec077d952f9e065915ecd361904f1dee2e2ecd662b4314e0283) head=0xdca4d1bafe6db46378754f5812403cceab9260f5f417c29b56362c3c4cde2aa6
2025-06-05T19:20:15.581718Z ERROR node{idx=0}: downloaders::headers: Header cannot be attached to known canonical chain error=mismatched parent hash: got 0x5e2d7fa3d0610f458b29c1ded089ae9e6192149403f5d3d8741b081841c2e6fb, expected 0xdca4d1bafe6db46378754f5812403cceab9260f5f417c29b56362c3c4cde2aa6 number=29 hash=0x33adc1d8563c7e7b63d0591977784a331635fa9e80cbb7458c3b715c55db49ad
2025-06-05T19:20:15.581739Z DEBUG node{idx=0}: downloaders::headers: Resetting headers downloader
2025-06-05T19:20:15.581834Z DEBUG node{idx=0}:ChainOrchestrator::poll: sync::stages::headers: Commencing sync tip=Hash(0x50370fd9a6b5bec077d952f9e065915ecd361904f1dee2e2ecd662b4314e0283) head=0xdca4d1bafe6db46378754f5812403cceab9260f5f417c29b56362c3c4cde2aa6
2025-06-05T19:20:15.581856Z ERROR node{idx=0}:ChainOrchestrator::poll: sync::stages::headers: Cannot attach header to head error=mismatched parent hash: got 0x5e2d7fa3d0610f458b29c1ded089ae9e6192149403f5d3d8741b081841c2e6fb, expected 0xdca4d1bafe6db46378754f5812403cceab9260f5f417c29b56362c3c4cde2aa6
2025-06-05T19:20:15.581875Z  WARN node{idx=0}:ChainOrchestrator::poll: sync::pipeline: Stage encountered detached head stage=Headers local_head=BlockWithParent { parent: 0x14ac3021d584583ec913cdac45b5284d23998f1be27d8322eadde7ccfb24ab80, block: NumHash { number: 28, hash: 0xdca4d1bafe6db46378754f5812403cceab9260f5f417c29b56362c3c4cde2aa6 } } header=BlockWithParent { parent: 0x5e2d7fa3d0610f458b29c1ded089ae9e6192149403f5d3d8741b081841c2e6fb, block: NumHash { number: 29, hash: 0x33adc1d8563c7e7b63d0591977784a331635fa9e80cbb7458c3b715c55db49ad } } error=mismatched parent hash: got 0x5e2d7fa3d0610f458b29c1ded089ae9e6192149403f5d3d8741b081841c2e6fb, expected 0xdca4d1bafe6db46378754f5812403cceab9260f5f417c29b56362c3c4cde2aa6
2025-06-05T19:20:15.582040Z  INFO node{idx=0}:ChainOrchestrator::poll:Unwinding{stage=Finish}: sync::pipeline: Starting unwind from=28 to=25 bad_block=Some(28)
...
2025-06-05T19:20:15.748618Z  INFO node{idx=0}: reth_node_events::node: Preparing stage pipeline_stages=1/12 stage=Headers checkpoint=25 target=None
2025-06-05T19:20:15.748626Z ERROR node{idx=0}:ChainOrchestrator::poll: sync::stages::headers: Cannot attach header to head error=mismatched parent hash: got 0x5e2d7fa3d0610f458b29c1ded089ae9e6192149403f5d3d8741b081841c2e6fb, expected 0xdca4d1bafe6db46378754f5812403cceab9260f5f417c29b56362c3c4cde2aa6
2025-06-05T19:20:15.748644Z  WARN node{idx=0}:ChainOrchestrator::poll: sync::pipeline: Stage encountered detached head stage=Headers local_head=BlockWithParent { parent: 0x14ac3021d584583ec913cdac45b5284d23998f1be27d8322eadde7ccfb24ab80, block: NumHash { number: 28, hash: 0xdca4d1bafe6db46378754f5812403cceab9260f5f417c29b56362c3c4cde2aa6 } } header=BlockWithParent { parent: 0x5e2d7fa3d0610f458b29c1ded089ae9e6192149403f5d3d8741b081841c2e6fb, block: NumHash { number: 29, hash: 0x33adc1d8563c7e7b63d0591977784a331635fa9e80cbb7458c3b715c55db49ad } } error=mismatched parent hash: got 0x5e2d7fa3d0610f458b29c1ded089ae9e6192149403f5d3d8741b081841c2e6fb, expected 0xdca4d1bafe6db46378754f5812403cceab9260f5f417c29b56362c3c4cde2aa6
2025-06-05T19:20:15.748703Z  INFO node{idx=0}:ChainOrchestrator::poll:Unwinding{stage=Finish}: sync::pipeline: Starting unwind from=25 to=25 bad_block=Some(28)

here the problematic part is Starting unwind from=25 to=25 bad_block=Some(28) which only happened because of HeaderStage receiving message only relevant for the previous run