* use identical progress counter

* fix gc total and valid count

* docs: add docs for `stats` command

* docs: set `--interval 0` in `profile` command to make replay finished as
fast as possible

* cmd: add `stats` command to show runtime stats

* use pre-defined schema

* add metric for used buffer size

* add blockcache, and adjust latency color

* add meta ops histogram

* fix display of large latency

* change meta ops name

* use verbosity level instead of detail flag, so we may add more items
later

* print dimmed value between intervals

an immediate result

* build tikv engine by default

* address comments

When writeback-cache mode is enabled for FUSE, it is possible to READ a file opened with O_WRONLY.
Refs: https://www.kernel.org/doc/Documentation/filesystems/fuse-io.txt

* meta: when creating an existed file, open it instead

* fix attr

* fs: use exclusive create

* fix check type

* fix when attr is nil

* fix import

* move prefix out of kvMeta

* fix npe

* feat: progress bar for dump and load

* refactor: move the code of create the progress bar into pkg/utils/utils.go

* refactor: move the code of create the progress bar into pkg/utils/utils.go

* feat: progress bar for dump and load,move the code of create the progress bar into pkg/utils/utils.go

* add show progress status for list slice

* Change showProgress type to function

* cleanup compaction

* fix npe

* add rwlock to read/write file

* enable fsx test

* lock the handle in truncate()

* restart short read

* use smaller file

* fix truncate

* fix truncate and integration tests

* usedSpace

* cleanup

* disable fsx test temporary

* add ToC for command reference Chinese version

* fix link

* fix link

* add ToC for command reference English version

* avoid the memory copy in cache

* fix bug

* fix npe

* docs: add docs for TiKV as metadata engine

* add juicefs version for benchmark results

* Fix doc

* Minor edit

* add tikv information
Co-authored-by: NChangjian Gao <gcj@juicedata.io>

* hadoop: update

* hadoop: remove show cmd

* hadoop: rename .jfsconfig to .config

* hadoop: make SecretKey and EncryptKey invisible

* cleanup
Co-authored-by: NDavies Liu <davies@juicedata.io>

* alloc memory from buffer pool

* 1. slow down write when there are more than 1000 pending slices
2. flush the slices earlier when there are more than 800
3. stop compacting when use 150% of buffer size
4. increase limit of compaction to 1000 slices
5. increase the trigger to compaction to 100 slices