perf(instrumentation-pg): reduce temp objects allocations

[Samuron]

Which problem is this PR solving?

The amount of allocations that opentelemetry produces is too high

Short description of the changes

Replaced attributes passed on span creation with span.setAttirbutes to avoid multiple remapping inside startSpan function, reused kind by providing constant and replaced string interpolation with constants

[Flarna]

note that adding attributes after startSpan has the side effect that they are not available for a sampler.
Therefore a sampling decision based on e.g. SemanticAttributes.DB_SYSTEM is no longer possible after this change.

[Samuron]

@Flarna oh, makes sense. Guess all is left is to reduce one temp allocation with spreading then, I've updated the MR

[blumamir]

Thank you for the contribution and for help improving the code.
Added 2 nit comments

[codecov]

Codecov Report

Merging #2019 (53c2f8f) into main (dfb2dff) will increase coverage by 4.70%.
Report is 15 commits behind head on main.
The diff coverage is n/a.

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #2019      +/-   ##
==========================================
+ Coverage   90.97%   95.68%   +4.70%     
==========================================
  Files         146       11     -135     
  Lines        7492      788    -6704     
  Branches     1502      164    -1338     
==========================================
- Hits         6816      754    -6062     
+ Misses        676       34     -642     

see 136 files with indirect coverage changes

[blumamir]

Thanks for addressing the previous comments so quickly 🥇
added one last nit comment about the index.ts