Correct a debug assertion

Previously, the 8-byte nonce after each mini-transaction was not covered
by the mini-transaction checksum. That could cause catastrophic failure
when the server was killed at a time where the write of the checksum
is durable but the nonce is not.

This changes the encrypted format in two ways: The nonce will be written
before the checksum, and it will be included in the checksum.

Using neither flush_lock nor write_lock around PMEM writes seems
to yield the best performance. It still is somewhat slower than
the pwrite() and fdatasync() based interface. Maybe some CLDEMOTE
in mtr_t::commit() would help?

log_close(): Update log_sys.write_to_buf.

mtr_t::commit(): Remove the conditional update of log_sys.write_to_buf.
We do write to the log even if we are not marking pages modified for
the first time.

mtr_t::commit_shrink(): Let log_close() in mtr_t::finish_write() update
log_sys.write_to_buf.

A quick test with the log on a PCIe NVMe drive yields the best
throughput without O_DIRECT. With O_DIRECT|O_DSYNC
(enabled by innodb_flush_method=O_DSYNC) both the average and
the maximum throughput are a little faster.

The slowest would be O_DIRECT without O_DSYNC.
Hence, it probably is better to not use O_DIRECT on the log file
when not also setting O_DSYNC.

mtr_t::finish_write(): Avoid invoking
log_sys.set_check_flush_or_checkpoint() if log_sys.is_pmem().
The log_sys.max_buf_free has no meaning on the mmap-based log.

With this, the mmap-based log writes are finally faster than the
pwrite() and fdatasync() based writes.

One PMEM device would deliver about 800 transaction commits per second
with pmem_deep_persist() and about 127,000 commits per second with
pmem_persist().

We will basically assume that the PMEM device does not lose writes
on power loss, just like the previous interface
pmem_memcpy_persist() did.

Invoking pmem_deep_persist() turns out to make things very slow
on PMEM. It is better to always use flush_lock for group commit.

Actually use group commit in trx_flush_log_if_needed_low()