Fix SQLite session_id

[S7evinK]

This fixes an issue with device_id/session_ids.
If a device_id is reused, we would reuse the same session_id, since we delete one device and insert a new one directly, resulting in the query to get a new session_id to return the previous session_id. (SELECT count(access_token))

For reference: matrix-org/complement#613

[codecov]

Codecov Report

Base: 63.27% // Head: 63.25% // Decreases project coverage by -0.03% ⚠️

Coverage data is based on head (15ea20c) compared to base (11d9b9d).
Patch coverage: 68.62% of modified lines in pull request are covered.

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #2977      +/-   ##
==========================================
- Coverage   63.27%   63.25%   -0.03%     
==========================================
  Files         499      499              
  Lines       53612    53651      +39     
==========================================
+ Hits        33924    33937      +13     
- Misses      15992    16011      +19     
- Partials     3696     3703       +7     

Flag	Coverage Δ
unittests	42.24% <68.62%> (-0.03%)	⬇️

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

Impacted Files	Coverage Δ
userapi/storage/postgres/devices_table.go	85.44% <25.00%> (-1.10%)	⬇️
userapi/storage/shared/storage.go	77.55% <70.00%> (-0.25%)	⬇️
userapi/storage/sqlite3/devices_table.go	84.21% <76.47%> (-0.37%)	⬇️
internal/caching/cache_serverkeys.go	53.84% <0.00%> (-46.16%)	⬇️
federationapi/storage/shared/storage.go	75.30% <0.00%> (-1.86%)	⬇️
mediaapi/routing/upload.go	56.19% <0.00%> (-1.77%)	⬇️
roomserver/storage/sqlite3/rooms_table.go	74.31% <0.00%> (-1.64%)	⬇️
roomserver/internal/query/query.go	55.17% <0.00%> (-0.44%)	⬇️
roomserver/storage/shared/storage.go	65.86% <0.00%> (ø)
syncapi/storage/shared/storage_sync.go	61.92% <0.00%> (ø)
... and 3 more

Help us with your feedback. Take ten seconds to tell us how you rate us. Have a feature suggestion? Share it here.

☔ View full report at Codecov.
📢 Do you have feedback about the report comment? Let us know in this issue.

[hughns]

For future reference, the complement test ended up in matrix-org/complement#622