feat(updates): add getDifference reason and unknown-peer diagnostics

[ernado]

Motivation

Reports like #1623 ("some UpdateNewMessage updates are not received until I send a message myself") are hard to diagnose from logs. The updates package already logs gap detection in detail, but two crucial things were invisible:

1. Why a getDifference fired — it is called from ~20 sites (startup, pts/qts/seq gap timeouts, idle timeout, the access-hash peer gates from feat(updates): force getDifference for updates referencing peers with unknown access hash #1738, updatesTooLong, seq-pts-changed, slice/too-long continuations, channel subscribe/too-long/idle) and every one logged the same bare "Getting difference".
2. Which peer was unresolved — the feat(updates): force getDifference for updates referencing peers with unknown access hash #1738 access-hash gate (userPeersKnown) silently forced a getDifference with no record of the offending user, which is the exact signature of the "stranger messages me first" case.

Changes

• Thread a reason string through getDifference/getDifferenceLogger for both the common (internalState) and channel (channelState) paths. The "Getting difference" log now includes the reason plus current pts/qts/seq/date (channel_id/pts for channels).
• Log the offending user_id (and whether it was a hasher error vs. a cache miss) in userPeersKnown when an access hash is unknown.

A #1623-style trace now reads:

User access hash unknown, forcing getDifference  user_id=12345 hasher_error=false
Getting difference  reason=short-message-peer-access-hash-unknown pts=9001 qts=1 seq=42 date=...

reason values are stable kebab-case tokens (pts-gap-timeout, combined-peer-access-hash-unknown, channel-too-long, …) so they are greppable and could later back a metric label.

Notes

• Diagnostics only — no behavior change to the state machine.
• All new logging is at Debug level (one existing Error path also gets the reason), consistent with the package style; zero noise unless a debug logger is wired in.

Testing

• go test ./telegram/updates/ passes
• go build ./..., go vet, gofmt clean

🤖 Generated with Claude Code

[codecov]

Codecov Report

❌ Patch coverage is 59.57447% with 19 lines in your changes missing coverage. Please review.
✅ Project coverage is 71.39%. Comparing base (4793972) to head (44f77ce).

Files with missing lines	Patch %	Lines
telegram/updates/state.go	64.00%	8 Missing and 1 partial ⚠️
telegram/updates/state_channel.go	43.75%	9 Missing ⚠️
telegram/updates/state_apply.go	0.00%	1 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #1781      +/-   ##
==========================================
+ Coverage   71.32%   71.39%   +0.06%     
==========================================
  Files         504      504              
  Lines       23625    23640      +15     
==========================================
+ Hits        16851    16878      +27     
+ Misses       5542     5533       -9     
+ Partials     1232     1229       -3     

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.