Reduce excess debug logging, simplify logging implementation

[rustyrussell]

A quick audit of my node reveals the top logging occurrences:

    1589311 DEBUG   02e01367e1d7818a7e9a0e8a52badd5c32615e07568dbe0497b6a47f9bef89d6af-connectd: peer_out WIRE_WARNING
     139993 DEBUG   lightningd: fixup_scan: block 786151 with 1203 txs
      55388 DEBUG   plugin-bcli: Log pruned 1001 entries (mem 10508118 -> 10298662)
      33000 DEBUG   gossipd: Unreasonable timestamp in 0102000a38ec41f9137a5a560dac6effbde059c12cb727344821cbdd4ef46964a4791a0f67cd997499a6062fc8b4284bf1b47a91541fd0e65129505f02e4d08542b16fe28c0ab6f1b372c1a6a246ae63f74f931e8365e15a089c68d61900000000000d9d56000ba40001690fe262010100900000000000000001000003e8000001f30000000000989680
      23515 DEBUG   hsmd: Client: Received message 14 from client
      22269 DEBUG   024b9a1fa8e006f1e3937f65f66c408e6da8e1ca728ea43222a7381df1cc449605-hsmd: Got WIRE_HSMD_ECDH_REQ
      14409 DEBUG   gossipd: Enqueueing update for announce 0102002f7e4b4deb19947c67292e70cb22f7fac837fa9ee6269393f3c513d0431d52672e7387625856c19299cfd584e1a3f39e0f98df13c99090df9f4d5cca8446776fe28c0ab6f1b372c1a6a246ae63f74f931e8365e15a089c68d61900000000000e216b0008050001692e1c390101009000000000000003e800000000000013880000004526945a00
      12534 DEBUG   gossipd: Previously-rejected announce for 514127x248x1
      12092 DEBUG   connectd: Bad cupdate for 641641x1164x1/1, ignoring (delta=80, fee=1073742199/58)
      10761 DEBUG   02e01367e1d7818a7e9a0e8a52badd5c32615e07568dbe0497b6a47f9bef89d6af-channeld-chan#70770: Got it!
      10761 DEBUG   02e01367e1d7818a7e9a0e8a52badd5c32615e07568dbe0497b6a47f9bef89d6af-channeld-chan#70770: ... , awaiting 1120
      10761 DEBUG   02e01367e1d7818a7e9a0e8a52badd5c32615e07568dbe0497b6a47f9bef89d6af-channeld-chan#70770: Sending master 1020

This series seeks to address some of those (though the fixup one is transient).

[cdecker]

Nice changes, that log infrastructure needed some love :-)

ACK https://github.com/ElementsProject/lightning/pull/8770/commits/fad257ee2b3a178849f9d23b0e94f0da63a485b5

[cdecker]

The CI failures are true positives, since the getlog behavior has changed (the level tiering seems to have changed).