Stopping lbrynet fails on full disk

Nobody has claimed this yet.

## The Issue

Trying to stop the daemon when the disk is full fails.
Trying to stop it using systemd also leaves a single process running which can only be solved with kill -9
Lbrynet seems to be using disk space at runtime?

### Steps to reproduce
1. fill the disk
2. run brynet stop
3. when it fails, check disk space
4. forcefully terminate the process
5. watch disk space again

### Expected behaviour
The daemon stops

### Actual behaviour
./lbrynet stop
[15609] INTERNAL ERROR: cannot create temporary directory!

```
$ sudo service lbrynet stop
$ ps x
PID TTY STAT TIME COMMAND
...
14500 ? Sl 45:37 /opt/lbry/lbrynet start
15964 pts/0 R+ 0:00 ps x
```

## System Configuration
```
{
"id": null,
"jsonrpc": "2.0",
"result": {
"build": "release",
"desktop": "Unknown",
"distro": {
"codename": "Xenial Xerus",
"id": "ubuntu",
"like": "debian",
"version": "16.04",
"version_parts": {
"build_number": "",
"major": "16",
"minor": "04"
}
},
"lbrynet_version": "0.30.3",
"lbryschema_version": "0.0.16",
"os_release": "4.4.0-138-generic",
"os_system": "Linux",
"platform": "Linux-4.4.0-138-generic-x86_64-with-debian-stretch-sid",
"processor": "x86_64",
"python_version": "3.7.1"
}
}
```

## Anything Else
It seems that when I force stop the daemon, in some cases, a lot of space is freed up. Why is that?

```
[lbry@whatever:~] $ df -h
Filesystem Size Used Avail Use% Mounted on
/dev/vda1 78G 78G 0 100% /

[lbry@whatever:~] $ sudo service lbrynet stop

... out of the 2 lbrynet processes only 1 terminates, so one needs to be force killed
[lbry@whatever:~] $ kill -9 14500

[lbry@whatever:~] $ df -h
Filesystem Size Used Avail Use% Mounted on
/dev/vda1 78G 49G 29G 64% /

```

## Internal Use

### Acceptance Criteria
1.
2.
3.

### Definition of Done
- [ ] Tested against acceptance criteria
- [ ] Tested against the assumptions of user story
- [ ] The project builds without errors
- [ ] Unit tests are written and passing
- [ ] Tests on devices/browsers listed in the issue have passed
- [ ] QA performed & issues resolved
- [ ] Refactoring completed
- [ ] Any configuration or build changes documented
- [ ] Documentation updated
- [ ] Peer Code Review performed

Reproduce the failure by filling the disk, running `lbrynet stop`, and comparing the `service lbrynet stop` process list and disk usage before and after force-killing the remaining process. Trace the daemon shutdown path from the `lbrynet stop` and systemd/service entry points; done means the daemon stops cleanly on a full disk, no process remains, and regression coverage is added.

Written by the indexing model from the issue text.