Attempts to resolve issues since #3210 in Lua loading Other Devices and the underlying OTA issues which cause them when running in Wide switch mode at 1:16 ratio or slower.

Issue 1: Too many syncs / too short TlmBoost

Changes to tx_main.cpp. TlmBoost is correctly activating since the previous PR, but the logic that requests a SYNC packet to signal the RX to change ratios expected to go to 1:2 ratio, and therefore would request a ratio change for each piece of data it wanted to send. The extra sync packets would then immediately disable boost mode and set the ratio back to the user-configured value as soon as the upload was complete, leaving the downlink in slow mode. This caused the DEVICE_INFO packets from the RX to take 20-30s to arrive at 1:128, and even longer for any flight controller's DEVICE_INFO.

Issue 2: Lua expects all DEVICE_INFO to arrive before it finishes loading

Changes to elrsv3.lua. If you wait long enough, "Other Devices" will appear in the Lua, but clicking on it has 0 items in the list. This is because it only populates the list at the end of loading the main set of parameters, and the overly slowed remote info wouldn't arrive in time. I've rearranged the code so that the list is populated when the Other Devices folder is populated, as well as if a new DEVICE_INFO is received while in that folder.

The field re-request timeout has also been extended from 0.5s to 5.0s, as requesting every 0.5s at slow 50Hz 1:8 telemetry caused an abundance of repeated parameter chunks to be downloaded from the remote side.

These changes are not required as after the other fixes, the list almost always loads the Other Devices properly. However, loading of the RX's lua is roughly half the speed due to all the duplicated requests, and I believe the Other Devices list should more reliably populate.

Issue 3: Requesting next parameter before the previous has completed downloading

Changes to telemetry.cpp / lua.cpp. When stubborn_receiver completes a download from the RX, the data is forwarded to the lua as soon as the transfer is completed. However, the RX still has the telemetry slot locked until it receives the final ACK. With Wide switch mode, the telemetry ack bit is only sent every 8 channels packets, which has a 50/50 chance of falling on the telemetry slot, which causes it to get pushed to every 16 packets. Before that 16 packet timeframe, the lua will have requested the next parameter chunk. However, since the receiver still has the slot locked, there is nowhere to put it and the response is discarded. This leads to the loading stalling until a re-request of the parameter chunk.

The receiver code has been updated to retry queueing the parameter every 4 packets until it is successful, or another request has come in (which will replace the pending request).

New Behavior

If the RX ever queues a PARAM_INFO message all other telemetry is now paused (dropped) for 10 seconds. During this time, only DEVICE_INFO and PARAM_INFO type messages will be allowed into the queue. This is to speed up the loading of the remote lua while under load coming from a connected flight controller (such as rotorflight which does a good job of using the entirety of the downstream bandwidth). While the remote lua is loading, users will experience "sensor lost" messages from EdgeTX.

Bugfixes

• There was a stack overrun issue in sendCRSFparam() on the RX where the buffer (paramInformation) for the is only DEVICE_INFORMATION_LENGTH (~32 bytes). This can not hold the full 64 bytes as needed for a full parameter chunk.
• telemetry.cpp allocates a 64 byte slot for DEVICE_INFO which is not used. All DEVICE_INFO packets go into the 2-slot slots, with the RX forced to use slot 2, and the Flight Controller using whichever slot is open. The extra buffer has been removed.
• Telemetry on the RX was using overlapping buffer areas for its queue, leading to the 2-slot buffer actually reusing the fixed slots. This caused corruption of DEVICE_INFO and PARAM_INFO packets from the RX, or corruption and dropping of standard telemetry items when the 2-slot was used. This overlap has been resolved.

Going mad with power

This also adds a version number to the lua so users can tell what they are running. It appears in the EXIT line and we'll have to manually tag it in the source file.