Add atomic register access for all backends

[dNechita]

PR Description

Local Backend

First patch introduces thread-safe register read/write operations for the local backend using direct_reg_access via debugfs. It adds two new backend ops:

• reg_read()
• reg_write()

These functions use file locking (flock) to ensure atomic access and prevent race conditions when multiple threads interact with the same register interface.

The fallback logic in iio_device_reg_read() and iio_device_reg_write() is preserved for backends that do not implement these atomic operations.

This improves reliability for multi-threaded applications using libiio with the local backend.

Network, USB and Serial Backends

Second patch implement atomic register operations which ensures that a register read or a register write cannot be interrupted by other clients of IIOD.
Advantages:

• half round trips: single command vs write-address + read-value
• binary addresses/values vs hex string conversion

On client-server compatibility:
v1.0 clients can connect to v1.0+ servers (atomic ops used)
v1.0 clients connecting to v0.x servers fall back to attribute method

Closes #1351.

PR Type

• Bug fix (a change that fixes an issue)
• New feature (a change that adds new functionality)
• Breaking change (a change that affects other repos or cause CIs to fail)

PR Checklist

• I have conducted a self-review of my own code changes
• I have commented new code, particularly complex or unclear areas
• I have checked that I did not introduce new warnings or errors (CI output)
• I have checked that components that use libiio did not get broken
• I have updated the documentation accordingly (GitHub Pages, READMEs, etc)

[rgetz]

Don't you want to fix this at the same time?

     * NOTE: This has a race condition when multiple clients access
     * the same IIOD server, as the two-step process (write address,
     * then read value) is not atomic.

[dNechita]

Don't you want to fix this at the same time?

     * NOTE: This has a race condition when multiple clients access
     * the same IIOD server, as the two-step process (write address,
     * then read value) is not atomic.

Yes, I think this can be removed. The race condition won't happen on IIOD server anymore.
Thanks!