Parchains scheduler: ttl value is off

[eskimor]

This check is nice, but in reality ttl should never be larger either as it then would allow one parachain to starve the other in a core sharing scenario.

We should just derive ttl from the lookahead directly and deprecate the ttl configuration.

The idea here was that collations can be backed out of order, but we don't actually do that and also agreed that it is a bad idea, with very little gain, but added complexity.

• Deprecate/remove ttl configuration value
• Ignore its value and instead make it always equal claim queue length
• Fix this check to <=.
• Test that backing in a core sharing fashion works, even if one para does not provide collations.
• Test that classic synchronous backing (12s block times) continues to work.

[sandreim]

Closing in favor of #4776 which removes the TTL completely.