Optimize performance bottlenecks in adaptive concurrency, storage, and scoring

[Copilot]

Identified and fixed inefficient algorithms and memory allocation patterns causing unnecessary O(n) operations and excessive GC pressure during scoreboard generation and participant management.

Changes

Fixed P95 percentile calculation in adaptive concurrency

• Before: Approximated as max * 0.8 (inaccurate)
• After: Proper sorting-based percentile calculation
• Impact: Accurate API concurrency adjustments under load

// Old: wrong percentile
cutoff := time.Duration(float64(maxTime) * 0.8)

// New: actual 95th percentile
sortedTimes := make([]time.Duration, len(window))
copy(sortedTimes, window)
sort.Slice(sortedTimes, func(i, j int) bool { return sortedTimes[i] < sortedTimes[j] })
return sortedTimes[int(float64(len(sortedTimes)) * 0.95)]

Optimized in-memory storage with O(1) name lookups

• Before: O(n) linear search through all participants for duplicate names
• After: Added participantsByName hash map index
• Impact: Name validation scales from O(n) to O(1)

Pre-allocated maps and slices

• scoring/calculator.go: Pre-size problem ID map with capacity hint
• storage/in_memory.go: Pre-allocate slice capacity in fetchStartingProblems
• Impact: Reduced reallocation overhead and GC pressure

Memory pooling for string builders

• bot/scoreboard.go: Reuse string builders from performance pool with capacity hints
• Impact: Eliminates repeated allocations during scoreboard formatting

Added HTTP status code constants

• Replaced magic numbers (404, 429, 500) with named constants
• Removed duplicate HTTPServerErrorThreshold definition

Original prompt

Identify and suggest improvements to slow or inefficient code

---

✨ Let Copilot coding agent set things up for you — coding agent works faster and does higher quality work when set up for your repo.