feat: support the Logitech G502 Lightspeed and G915

[PyFlat-JR]

What does this PR do?

Adds a second Logitech HID++ protocol for devices that report a battery voltage (0x1001) instead of a percentage, with two models on it: the G502 Lightspeed (receiver C539, cable C08D) and the G915 keyboard (receiver C547, cable C33E). Closes #21 and part of #25. The protocol is based on the changes @Luckylucasv shared in #21.

• Protocol: LogitechVoltageBatteryProtocol sits on LogitechHidppProtocol, the same as the existing mouse and headset protocols, at device index 1. Bytes 4-5 hold the voltage in mV, and byte 6 holds flags.
• Percentage: the mV are mapped through the Linux kernel's 100-entry 0x1001 table (hidpp20_map_battery_capacity), interpolated linearly. It's an estimate, since G HUB uses its own per-device table. PercentFromMillivolts moved into LogitechHidppProtocol, and the headset protocol uses it too.
• Flags: these follow the kernel and Solaar.
  • Bit 7 means external power. Only while it's set do bits 0-2 give the charge state: 0 charging, 1 full, 2 not charging.
  • Bit 5 is the critical-level bit, never "full".
  • 0 mV (off or asleep) throws, so the plugin keeps the last value instead of showing 0%.
• Charging: while charging there's no percentage, because the voltage reads high under charge, as on Rapoo, AULA and MCHOSE. Full reads 100%.
• G915 and the G502 X PLUS: they share the C547 receiver. The G915 has 0x1001 and no 0x1004, so the existing fallback rule from feat: support the Logitech G502 X PLUS #29 keeps each one on its own receiver.
• Hardware test: Reads_the_raw_logitech_battery_voltage prints the raw mV and flags next to the parsed reading, for calibrating against G HUB.
• Linux: udev lines for c539, c08d and c33e (c547 is already there), with the guide's copy kept in sync.
• Docs: README rows ("approx.", no level while charging) and the protocol in AGENTS.md.

Checklist

• I have read AGENTS.md and CONTRIBUTING.md.
• dotnet build and dotnet test pass locally (0 warnings, 343 tests).
• I updated README.md / AGENTS.md if this changes what they describe.
• If this touches a device backend: I tested it against the real hardware (see below).

Device/hardware tested against (if applicable)

G502 Lightspeed: tested by its owner (@Luckylucasv) in #21 on Windows, on an earlier state of this branch (before the full kernel table and the no-level-while-charging change):

• Receiver: 96% Discharging. Cable: Charging, with the percentage jumping to about 100.
• Compared with G HUB at 90-100%: 2-4% apart. One reading showed G HUB at 72% while the plugin showed 97%; that is not explained yet.
• Off: the last value stays, shown as stale. With the receiver and the cable both connected, the mouse appears once and reads from the cable.
• In Macro Deck 3 over 1-2 hours, the variables and widget behaved.

G915: tested by its owner (@ReqwiemG) in #25 on Windows, with the G502 X PLUS and PRO X 2 LIGHTSPEED on their own receivers and G HUB running:

• Receiver, with all receivers and no cables: 80% Discharging (3990 mV) against G HUB's 82%. The G502 X PLUS still read correctly on its own receiver.
• Cable: Charging (4042 mV, flags 0x80) while G HUB showed 82% charging. The plugin estimated 87% then, which is why charging now shows no percentage.