fix: Block I/O and Network I/O monitoring charts not populating

[Siumauricio]

Fixes #4578

recordAdvancedStats stored the raw docker stats strings with units ("610kB", "938169.55MB") in readMb/writeMb/inputMb/outputMb, and the recharts area charts received them as numeric dataKeys → NaN → empty charts. CPU/Memory worked because their charts already strip % / parse units.

• New monitoring/units.ts with toMb/parseIoToMb/formatMb (pure module, shared by server and client like hostname-validation)
• recordAdvancedStats now stores numeric MB for block/network, fixing both container-level and server-level (getHostSystemStats) monitoring
• Charts and headers parse legacy string data points already recorded on disk, so existing installs render immediately after upgrading
• Headers now format units properly (In: 938.65 GB instead of In MB: 938169.55MB)

No migrations.

Greptile Summary

This PR normalizes Docker block and network I/O statistics to numeric megabytes before persistence while retaining compatibility with legacy string values.

• Adds shared parsing and display-formatting utilities for decimal and binary I/O units.
• Converts newly recorded container and host I/O metrics to numeric MB values.
• Updates free-monitoring charts and headers to parse legacy data and display appropriate units.

Confidence Score: 5/5

The PR appears safe to merge with no concrete blocking or independently actionable non-blocking issues identified.

The new conversion helpers cover the units emitted by the repository's container and host monitoring paths, numeric values remain unchanged when rendered, and legacy persisted strings are parsed by both the charts and headers.

_{Reviews (1): Last reviewed commit: "fix: parse Block I/O and Network I/O sta..." | Re-trigger Greptile}

Context used:

• Knowledge Base — Monitoring and Live Terminal/Log Streaming