refactor!: harden verification

[Wondertan]

We realized that non-adjacent verification might produce false negatives in Tendermint consensus. Additionally, not all chains have non-adjacent verification, and before this change, we dropped all headers that failed the non-adjacency verification. However, we should give them a second chance and keep them as sync targets.

Additionally:

• Introduces typical verification flow. Users had to verify time, height, and chain-id themselves, but not anymore.
• We reject known headers, and it's a hard failure now.
• Non-adjacent verification failure is considered a soft failure.

Closes #79

[codecov-commenter]

Codecov Report

Merging #90 (84b76c3) into main (c5a5780) will increase coverage by 0.14%.
The diff coverage is 91.66%.

@@            Coverage Diff             @@
##             main      #90      +/-   ##
==========================================
+ Coverage   68.27%   68.41%   +0.14%     
==========================================
  Files          37       37              
  Lines        2909     2929      +20     
==========================================
+ Hits         1986     2004      +18     
- Misses        771      774       +3     
+ Partials      152      151       -1     

Files Changed	Coverage Δ
p2p/helpers.go	78.26% <ø> (ø)
p2p/server.go	77.22% <ø> (-5.00%)	⬇️
p2p/subscriber.go	44.64% <0.00%> (ø)
sync/sync_head.go	68.50% <80.00%> (+6.78%)	⬆️
verify.go	96.22% <96.22%> (ø)
headertest/dummy_header.go	72.85% <100.00%> (+5.14%)	⬆️
headertest/dummy_suite.go	100.00% <100.00%> (ø)

... and 4 files with indirect coverage changes

[vgonkivs]

LGTM