avoid exclusive lock during update_asset_storage

[blamelessgames]

noticed a coarsely held exclusive lock in update_asset_storage that was causing contention - occasionally one version or another of the wrapping system would be held for 10+ms in my profiling.

narrowed the lock scope to the point of use. if this solution seems too quick-and-dirty i'm willing to re-examine

[cart]

yeah asset lifecycle events should be pretty rare, so grabbing the lock on each event doesn't seem terrible. The tradeoff is that we get way more "lock-ey" when we load many assets at once. We could optimize that by adding a let asset_sources: Option<RwLockWriteGuard<HashMap<SourcePathId, SourceInfo>> = None right before the loop starts, then storing the guard in there, which should let us grab the guard once across all items processed.