Add exponential backoff to Fleet Desktop server communications

[sharon-fdm]

Closes #45624
Closes #45625

Summary

• Adds orbit/pkg/backoff package: stateful exponential backoff tracker with jitter, thread-safe, per-path isolation
• Integrates backoff into Fleet Desktop's ping loop and checkToken retry loop
• On error: interval doubles each failure (10s, 20s, 40s, 80s, ...) capped at 30 minutes
• On success: resets immediately to normal polling interval
• Each communication path tracks its own backoff independently

Addresses corrective action #3 from the #44816 postmortem: Fleet Desktop kept polling /device/{token}/desktop at full rate with expired tokens, overwhelming the DB. With backoff, 250 hosts in this state produce ~0.14 req/s instead of 25 req/s (180x reduction).

Manual testing performed

1. Unit tests (19 total, all pass)

go test ./orbit/pkg/backoff/ -v -count=1   # 19 tests, 0 failures
make lint-go-incremental                    # 0 issues

13 logic tests: exponential doubling, cap at max, jitter range, reset on success, per-path isolation, concurrent access, no-give-up, monotonically non-decreasing intervals.

3 real-time ticker tests: actual time.Ticker with wall-clock measurements verifying intervals grow during failures and reset on success, matching how Desktop uses the tracker.

3 HTTP server integration tests: real httptest.TLSServer toggled between 200/401/500 mid-test, plus connection-refused scenario. Verified full lifecycle:

Phase 1 (healthy):    50ms, 50ms, 50ms
Phase 2 (401 errors): 51ms -> 109ms -> 216ms -> 448ms -> 508ms (cap)
Phase 3 (recovery):   instant reset to 50ms
Phase 4 (500 errors): 106ms -> 217ms -> 428ms
Phase 5 (recovery):   back to 50ms

2. Live Fleet server testing

Started Fleet dev server (FLEET_MYSQL_PASSWORD=fleet ./build/fleet serve --dev --dev_license), enrolled orbit, then ran test binary against real /device/{token}/desktop endpoint.

Invalid token (simulating #44816 expired-token scenario):

Attempt 1: HTTP 401, next=1.013s, failures=1
Attempt 2: HTTP 401, next=2.184s, failures=2
Attempt 3: HTTP 401, next=4.063s, failures=3
Attempt 4: HTTP 401, next=8.086s, failures=4
Attempt 5: HTTP 401, next=10s,    failures=5  (capped)
Attempt 6: HTTP 401, next=10s,    failures=6  (stays capped)

Server killed mid-poll (simulating server outage):

Attempt 1: HTTP 401,      next=1.042s, failures=1
Attempt 2: NETWORK ERROR, next=2.139s, failures=2  <-- server killed here
Attempt 3: NETWORK ERROR, next=4.154s, failures=3
Attempt 4: NETWORK ERROR, next=8s,     failures=4  (capped)

Seamless transition from HTTP errors to network errors -- same backoff behavior.

Recovery after server restart:

After success: interval=500ms, in_backoff=false, failures=0  <-- instant reset

3. Build verification

go build ./orbit/cmd/desktop/   # compiles clean
go build ./orbit/cmd/orbit/     # compiles clean

Limitation

Fleet Desktop requires a GUI (systray) for full end-to-end testing with the tray icon. The backoff logic was tested via the test harness above against real Fleet endpoints. A manual QA test on a machine with a display should: kill orbit, watch Desktop logs for "ping failed, backing off" messages with increasing intervals, then restore orbit and verify "exiting backoff" appears.

Test plan

• go test ./orbit/pkg/backoff/ -v -- 19 tests pass
• go build ./orbit/cmd/desktop/ -- compiles clean
• make lint-go-incremental -- 0 issues
• Manual: tested against live Fleet server with invalid tokens (401 backoff verified)
• Manual: tested server-down scenario (network error backoff verified)
• Manual: tested recovery (instant reset to base interval verified)
• Manual QA: enroll orbit + desktop on a display machine, kill orbit, verify Desktop logs show increasing backoff intervals, restore orbit, verify reset

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 66.76%. Comparing base (4c29a7a) to head (8383051).
⚠️ Report is 127 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main   #45623      +/-   ##
==========================================
+ Coverage   66.73%   66.76%   +0.02%     
==========================================
  Files        2732     2740       +8     
  Lines      218551   219009     +458     
  Branches    10840    10840              
==========================================
+ Hits       145857   146211     +354     
- Misses      59479    59573      +94     
- Partials    13215    13225      +10     

Flag	Coverage Δ
backend	68.59% <100.00%> (+0.02%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 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.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.