Rationale

At present, it takes about 35s to start the node, and it takes 25 to 30s to load the transaction cache in the node start period.
And the bloom filter is used as the transaction cache actually.
So it is expected to shorten the node start time by loading the bloom filter faster.

Why should this feature exist?
1.To make the node start faster.

Implementation

1.The node writes the bloom filter to the disk when the kill -15 signal is received.
2.Load the bloom filter into the ram if the bloom filter file is existed and can be opened normally,
or build the transaction cache from recent 65536 blocks like before.

Additional information

kill -15 is required

Related PRs

• feat(db/trans-cache): optimize for bloomFilter initialization #5394
• chore(gitignore):ignore the .cache folder generated by test #5448
• fix(db/trans-cache):avoid dumping a wrong cached data #5491
• fix(db/trans-cache): avoid recovering a wrong cached data #5505
• feat(db/trans-cache): verify cached files with crc32c #5523