fix: transaction retries should not timeout

[olavloite]

Transactions that are retried because of an aborted transaction use the retry settings of the Rollback RPC. This ensures reasonable backoff values. It however also meant that transactions that are retried multiple times could exceed the total timeout of the retry settings, and that again would cause the Aborted error to propagate. This change sets the total timeout for transaction retries to 24 hours and disables any max attempts in the retry settings to prevent retries to fail because the deadline is exceeded.

Transactions can still fail with timeout errors if individual RPC invocations exceed the configured timeout of that RPC. This change only prevents timeouts from occurring because of repeated retries of an entire transaction.

Fixes #1008

[codecov]

Codecov Report

Merging #1009 (6e1d08f) into master (a95f6f8) will decrease coverage by 0.00%.
The diff coverage is 100.00%.

@@             Coverage Diff              @@
##             master    #1009      +/-   ##
============================================
- Coverage     85.12%   85.11%   -0.01%     
+ Complexity     2619     2618       -1     
============================================
  Files           154      154              
  Lines         14324    14329       +5     
  Branches       1334     1334              
============================================
+ Hits          12193    12196       +3     
- Misses         1569     1570       +1     
- Partials        562      563       +1     

Impacted Files	Coverage Δ	Complexity Δ
...a/com/google/cloud/spanner/SpannerRetryHelper.java	84.61% <100.00%> (+3.66%)	3.00 <3.00> (ø)
...ain/java/com/google/cloud/spanner/SessionPool.java	88.93% <0.00%> (-0.20%)	71.00% <0.00%> (-1.00%)

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update a95f6f8...6e1d08f. Read the comment docs.