fix: use actual memory usage excluding cache/buffers in monitoring

[CorentinMre]

Description

Fixes memory usage calculation in monitoring to reflect actual memory usage excluding cache/buffers.

Problem

The current implementation uses v.Used which includes Linux cache and buffers, leading to:

• Inflated memory usage percentages (often 80-90% on healthy systems)
• False positive alerts when memory is actually available
• Misleading monitoring metrics

Solution

Changed the calculation to use Total - Available which:

• Excludes cache/buffers that can be freed automatically
• Matches the behavior of free -h command
• Provides accurate memory pressure monitoring

Testing

Tested on VPS (1.928 GiB RAM):

• Before: Monitoring showed 300 MiB while free -h showed 1.33 GiB used
• After: Both report 1.33 GiB (the real memory usage)

In Code

• Replace v.Used with (Total - Available) calculation

[CorentinMre]

CPU Usage

Using cpu.Percent(0, false) provides instantaneous measurements which are:

• Unreliable and often showing 0% or incorrect values
• Not representative of actual CPU load
• Inconsistent with standard monitoring tools

Solutions

Changed to use cpu.Percent(time.Second, false) which:

• Observes CPU usage over 1 second for accurate measurement
• Provides reliable and consistent metrics
• Aligns with standard monitoring practices

Testing

CPU:

• Before: Often showing "3%-4% or inconsistent values with instantaneous measurement
• After: Accurate measurements matching system load (tested under various loads)