txpool: continue pool pruning even if a tx can't be found

[j-berman]

• Defensive change from return to continue to ensure pool pruning executes to completion.
• Ensure tx pool weight doesn't underflow.
• Will propose a fix in PR wallet2, RPC: Optimize RPC calls for periodic refresh from 3 down to 1 call #8076 that prevents the root cause of the "Failed to find tx_meta in txpool" error. Noticed the cause of the error while reviewing that PR and that PR seems well positioned to solve it.

[rbrunner7]

Looks good.

Looking forward to the revealing of the root cause for that meta finding problem!

[SChernykh]

This makes it print the error once per daemon launch, not once per tx_memory_pool::prune call like it was before. Any specific reason for this change? To copy the previous behavior, warned shouldn't be static and it should be initialized outside of the while loop.

[j-berman]

Once the cap is reached, it will print here incessantly. I assume it wasn't realized that incessant printing would occur here because normal usage hasn't breached the default cap to my knowledge.

This matches #6881

monero/src/cryptonote_core/tx_pool.cpp

Lines 1451 to 1459 in d52b7d0

	txpool_tx_meta_t meta;
	if (!m_blockchain.get_txpool_tx_meta(sorted_it->second, meta))
	{
	static bool warned = false;
	if (!warned)
	MERROR(" failed to find tx meta: " << sorted_it->second << " (will only print once)");
	warned = true;
	continue;
	}