cln-plugin: Add dynamic configs and a callback for changes

[daywalker90]

I want this in cln-plugin: #7289

So this is my take for cln-plugin on it. The formatting changes are because @cdecker is not using rustfmt like a good boy and i noticed it too late, sorry. Some things get a bit weird here and there but overall it works!

I wanted existing projects to not have to do any changes if they were using static options (so all of them up to now), that's why i added the dynamic switch as a function, instead of an extra argument to the new_* functions.

Let me know what you think.

[cdecker]

That's me alright, bad boy for life 🤣
Will review ASAP

[cdecker]

Very nice change, just a minor suggestion to make dynamic() return a copy so we can just chain the invocations as if it were a Builder.

ACK 952da34

[cdecker]

ACK ac263e1

[daywalker90]

btw CLN printing broken logs if the plugin does not like the new option value and returns an Err() as a Result is a bit much imo:

**BROKEN** jsonrpc#70: Invalid setconfig summars-sort-by response from plugin: {"error":"Not a valid column name: `1`. Must be one of: GRAPH_SATS, OUT_SATS, IN_SATS, SCID, MAX_HTLC, FLAG, PRIVATE, OFFLINE, BASE, PPM, ALIAS, PEER_ID, UPTIME, HTLCS, STATE","id":"-c:setconfig#18/cln:setconfig#51","jsonrpc":"2.0"}

should be UNUSUAL, no?

[daywalker90]

and on the console you only get a generic error instead of the one from the plugin:

{
   "code": -1,
   "message": "Malformed setvalue summars-sort-by plugin return"
}

You have to then look at the broken log entry to get the error reported by the plugin. This could be improved!

[daywalker90]

So after the meeting and a private coding session with rusty it turned out that it works as it should with the python framework. So this is a skill issue on my part or an actual issue with the rust framework.
Closest i came to a well formed error:

pub async fn setconfig_callback(
    plugin: Plugin<PluginState>,
    args: serde_json::Value,
) -> Result<serde_json::Value, Error> {
      return Err(anyhow!(json!(RpcError {
          code: Some(-1),
          message: "bad thing".to_string(),
          data: None
      })));

**BROKEN** jsonrpc#68: Invalid setconfig clnrod-blockmode response from plugin: {"error":"{\"code\":-1,\"data\":null,\"message\":\"bad thing\"}","id":"pytest:setconfig#4/cln:setconfig#45","jsonrpc":"2.0"}

[daywalker90]

Ok , so cln-plugin was escaping RpcError's too much. Looking at the io log of an error coming from a plugin:

{"error":"{\"code\":-1,\"data\":null,\"message\":\"could not parse BlockMode from test\"}","id":"pytest:setconfig#4/cln:setconfig#45","jsonrpc":"2.0"}

I've added a commit where error messages get the proper (non-) escaping string. With that we get:

{"error":{"code":-1,"data":null,"message":"could not parse BlockMode from test"},"id":"pytest:setconfig#4/cln:setconfig#45","jsonrpc":"2.0"}