findAll: defer slips initialization (~20MB RAM savings)

[christopherkmoore]

Summary

• Defers slips.storages:map() from module load time to first search() call
• Prevents ~20MB of res.items entries from materializing into memory at addon load
• Result is cached after first search, so subsequent searches have no additional cost

Problem

At module level (line 195-198), findAll iterates every storage slip ID and accesses res.items[id] for each to build merged_storages_orders. This forces Windower's lazy-loading resource system to materialize hundreds of item entries into memory (~20MB).

This data is only needed when performing a search, not at idle.

Solution

Wrap the slips order initialization in ensure_merged_storages_orders(), called at the top of search(). Built once on first search, cached for subsequent searches.

Impact

• Idle memory: ~27MB → ~700KB
• First search: ~235ms one-time cost for cold res.items lookups (imperceptible)
• Subsequent searches: no change
• Multiboxers (6+ clients): saves ~150MB+ total RAM

Testing

• Verified //findall searches return correct results across all storage types including slips
• Verified tracker functionality unaffected
• Benchmarked cold res.items[id] lookups at ~0.2ms each (1000 lookups in 235ms)

[christopherkmoore]

I didn't do too much testing on this other than works on my machine. Claude helped me clean this up and it seems sensible with much better memory footprint after his change -- so I'll leave it up to you all to decide.