[MRG] Avoid changing precision in the backend

[kachayev]

Types of changes

Fix previously added (see PR #507) precision change when computing norm on the PyTorch backend.

Motivation and context / Related issue

Precision change is not necessary, and fails on devices which don't have support for double precision.

Related discussion #570.

How has this been tested (if it applies)

n/a

PR checklist

• I have read the CONTRIBUTING document.
• The documentation is up-to-date with the changes I made (check build artifacts).
• All tests passed, and additional code has been covered with new tests.
• I have added the PR and Issue fix to the RELEASES.md file.

[codecov]

Codecov Report

Merging #572 (7dc0579) into master (a56e1b2) will not change coverage.
The diff coverage is 100.00%.

Additional details and impacted files

@@           Coverage Diff           @@
##           master     #572   +/-   ##
=======================================
  Coverage   96.51%   96.51%           
=======================================
  Files          67       67           
  Lines       14821    14821           
=======================================
  Hits        14304    14304           
  Misses        517      517