ADBDEV-6599: Make gprestore --resize-cluster use --jobs for parallel restoration

[RekGRpth]

Make gprestore --resize-cluster use --jobs for parallel restoration

Users consider the current behavior to be incorrect, since the --jobs parameter
is used when creating a backup, which is not actually used during restoration.
Restoration works in a single instance of the gpbackup_helper agent (in the
--restore-agent mode).

In short, when restoring a backup using the gprestore utility with the
--resize-cluster option, which was previously made in parallel creation mode
(--jobs), the utility does not use the corresponding mode during restoration.
In this case, restoration occurs through gpbackup_helper, which is launched in
one instance of this process for each segment. Thus, one table is processed at
a time based on the list of table OIDs passed when gpbackup_helper is launched.

As a solution to the problem, this patch implements the launch of several
gpbackup_helper instances based on the value passed in the --jobs argument.

To do this:

1. The general list of table OIDs for restoration is distributed between
   several instances of gpbackup_helper processes according to the specified value
   of the --jobs parameter. Also this list is distributed between goroutines in
   main process by same way.

2. Since the general list of table OIDs for restoration is sorted by size, the
   distribution among gpbackup_helper instances and among goroutines is done
   according to the principle of sequentially distributing one table OID to each
   instance and goroutine. Thus, the load is expected to be more uniform. All
   batches corresponding to the same table OID go to the same gpbackup_helper
   instance and same goroutine.

3. An integer identifier of the "instance number" is added to the file name of
   the list of table OIDs for restoration by gpbackup_helper instances. They are
   not logically related in any way, the main thing is that its own list is passed
   as an argument when starting gpbackup_helper.

4. The instance number is also added to the name of the script file
   (scriptFile).

5. The instance number is also added to the name of the pipeFile file, error
   file and skip file.

6. To create the initial list of pipes, one pipe is created for the first table
   OID (according to the sorting of table OIDs by decreasing table size), which
   will be processed by each instance of gpbackup_helper. Thus, at the time of
   starting each instance of gpbackup_helper, the file system will already have
   the first pipe corresponding to the first table OID for this instance.

7. When starting gpbackup_helper, the value of the --jobs parameter is taken
   into account (in fact, the number of connections in the pool). The number of
   launched instances corresponds to the value of the parameter. The value of the
   path of the first pipe (pipeFile), the path to the agent startup script
   (scriptFile), and the path to the list of table OIDs corresponding to this
   instance are passed as arguments.

8. When deleting auxiliary files in DoCleanup, the files with the list of table
   OIDs and the startup script files (scriptFile) are deleted.

9. For --single-data-file and --copy-queue-size modes, the current behavior
   remains unchanged.

For more convenient management, all auxiliary files have been renamed. The
process pid is placed before the suffix. The word pipe has been removed from
the names of error and skip files.

New tests have been added and old ones have been adapted.

---

It is easier to view the changes with the "Hide whitespace" option enabled.

[silent-observer]

The new test has failed, please look into this

[RekGRpth]

Shouldn't this error file name be updated as well?

yes, added