Add transaction log

[whilo]

SUMMARY

Fixes #718

Checks

Bugfix

• Related issues linked using fixes #number
• Integration tests added
• Architecture Decision Record added if design changes necessary
• Formatting checked

Feature

• Implements an existing feature request. Make sure the feature request has been accepted for implementation before opening a PR.
• Related issues linked using fixes #number
• Integration tests added
• Documentation added
• Architecture Decision Record added
• Formatting checked

ADDITIONAL INFORMATION

[whilo]

For now we will stick to no transaction log to optimize for distributed readers and minimal IO during reading. The problem with transaction logs in our case would be that distributed readers would have to reapply them on every read, a disadvantage we also had with the hitchhiker-tree (in a fractal form). Since we have a single writer architecture and write throughput is generally fast enough such that we don't increase latency by writing each new path into the tree indices in parallel on a transaction, the main price we pay is more storage requirement and higher write throughput. The storage requirement can be handled with running gc regularly. And the write throughput is not that much better with a transaction log, too. As soon as you hit a high enough throughput rate and the log overflows on most transactions you have the same costs. So for now let's postpone the transaction log for simplicity and optimal read performance. On high write throughput Datahike also auto-batches transactions and reduces write amplification this way.