feat: adaptive resolution algorithm for chart data

Replaces the fixed period-to-aggregate mapping with a data-driven algorithm
that calculates optimal resolution per sensor based on actual reading frequency:

- High frequency sensors (>30/h): uses continuous aggregates (hourly/daily/monthly)
- Low frequency sensors (<~4/h): returns raw data without aggregation
- Boolean devices (REGANDO, EN COLA): returns state transitions + on/off percentage

The algorithm preserves diurnal patterns (e.g. temperature 9°C night → 23°C day)
by using hourly resolution for WEEK/MONTH scales on sensors that have enough data.

New StatsDao methods: getReadingDensity(), getRawReadings()
New DTO fields: resolution, pointCount, isBooleanDevice, transitions, booleanStats

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

Author: PabloHurtadoGonzalo86