fix: cancel inflight requests if database is canceled

[nolanlawson]

If database.close() is called, we should cancel any inflight IDB/fetch operations. This should eventually allow us to start setting up the database sooner than a microtask, since we can cancel the request if the user decides to change the data source.

[github-actions]

📊 Tachometer Benchmark Results

Summary

benchmark-total

• emoji-picker-element-change-tab: unsure 🔍 -1% - +1% (-0.17ms - +0.21ms)
  ^{this-change vs tip-of-tree}
• emoji-picker-element-database-interactions: unsure 🔍 -7% - +4% (-6.87ms - +3.79ms)
  ^{this-change vs tip-of-tree}
• emoji-picker-element-first-load: unsure 🔍 -5% - +4% (-2.38ms - +1.92ms)
  ^{this-change vs tip-of-tree}
• emoji-picker-element-search: unsure 🔍 -0% - +0% (-0.05ms - +0.06ms)
  ^{this-change vs tip-of-tree}
• emoji-picker-element-second-load: unsure 🔍 -7% - +4% (-2.80ms - +1.40ms)
  ^{this-change vs tip-of-tree}

Results

emoji-picker-element-change-tab

• Browser: chrome-headless 122.0.6261.111
• Sample size: 50
• Built by: Benchmarks #406

• Commit: 8062511

Version	Avg time	vs this-change	vs tip-of-tree tip-of-tree
this-change	33.33ms - 33.63ms	-	unsure 🔍 -1% - +1% -0.17ms - +0.21ms
tip-of-tree tip-of-tree	33.35ms - 33.58ms	unsure 🔍 -1% - +1% -0.21ms - +0.17ms	-

emoji-picker-element-database-interactions

• Browser: chrome-headless 122.0.6261.111
• Sample size: 50
• Built by: Benchmarks #406

• Commit: 8062511

Version	Avg time	vs this-change	vs tip-of-tree tip-of-tree
this-change	96.89ms - 104.23ms	-	unsure 🔍 -7% - +4% -6.87ms - +3.79ms
tip-of-tree tip-of-tree	98.23ms - 105.97ms	unsure 🔍 -4% - +7% -3.79ms - +6.87ms	-

emoji-picker-element-first-load

• Browser: chrome-headless 122.0.6261.111
• Sample size: 50
• Built by: Benchmarks #406

• Commit: 8062511

Version	Avg time	vs this-change	vs tip-of-tree tip-of-tree
this-change	50.38ms - 53.53ms	-	unsure 🔍 -5% - +4% -2.38ms - +1.92ms
tip-of-tree tip-of-tree	50.72ms - 53.65ms	unsure 🔍 -4% - +5% -1.92ms - +2.38ms	-

emoji-picker-element-search

• Browser: chrome-headless 122.0.6261.111
• Sample size: 50
• Built by: Benchmarks #406

• Commit: 8062511

Version	Avg time	vs this-change	vs tip-of-tree tip-of-tree
this-change	50.00ms - 50.06ms	-	unsure 🔍 -0% - +0% -0.05ms - +0.06ms
tip-of-tree tip-of-tree	49.98ms - 50.06ms	unsure 🔍 -0% - +0% -0.06ms - +0.05ms	-

emoji-picker-element-second-load

• Browser: chrome-headless 122.0.6261.111
• Sample size: 50
• Built by: Benchmarks #406

• Commit: 8062511

Version	Avg time	vs this-change	vs tip-of-tree tip-of-tree
this-change	34.90ms - 37.89ms	-	unsure 🔍 -7% - +4% -2.80ms - +1.40ms
tip-of-tree tip-of-tree	35.62ms - 38.57ms	unsure 🔍 -4% - +8% -1.40ms - +2.80ms	-

_{tachometer-reporter-action v2 for Benchmarks}

[github-actions]

Size Change: +420 B (+1%)

Total Size: 37.1 kB

Filename	Size	Change
./bundle.js	37.1 kB	+420 B (+1%)

_{compressed-size-action}

[nolanlawson]

I'm not sure I can actually get this to work properly. It's so much added complexity, and it still doesn't really improve the perf by doing what I'd really like to do, which is fire off the initial HTTP request before rendering.

It's a nice idea, but it's probably too complex to implement.