Avoid extra number conversions

[kappa]

We don't need BigFloat magic here as Grains.pm will always return bigints.

This should fix #239 and #240.

[kappa]

Ok, I am reading #240 comments and see that this is not the correct way to fix tests for exercises. I'll fix the pull request.

[kappa]

Travis failure fixed in #343.

[m-dango]

I'm not too sure what the problem we're solving is here. If bigint/bignum is removed from Grains.pm, this is the result of the test:

exercises/grains/.meta/solutions/grains.t .. 
# Seeded srand with seed '20200503' from local date.
1..12
ok 1 - Imported symbols: grains_on_square, total_grains
ok 2 - square no. 1
ok 3 - square no. 2
ok 4 - square no. 3
ok 5 - square no. 4
ok 6 - square no. 16
ok 7 - square no. 32
not ok 8 - square no. 64

# Failed test 'square no. 64'
# at exercises/grains/.meta/solutions/grains.t line 17.
# +---------------------+----+---------------------+-----+
# | GOT                 | OP | CHECK               | LNs |
# +---------------------+----+---------------------+-----+
# | 9223372036854779904 | == | 9223372036854775808 | 29  |
# +---------------------+----+---------------------+-----+
ok 9 - square 0 raises an exception
ok 10 - negative square raises an exception
ok 11 - square greater than 64 raises an exception
not ok 12 - returns the total number of grains on the board

# Failed test 'returns the total number of grains on the board'
# at exercises/grains/.meta/solutions/grains.t line 17.
# +----------------------+----+----------------------+-----+
# | GOT                  | OP | CHECK                | LNs |
# +----------------------+----+----------------------+-----+
# | 1.84467440737096e+19 | == | 18446744073709551615 | 36  |
# +----------------------+----+----------------------+-----+
Dubious, test returned 2 (wstat 512, 0x200)
Failed 2/12 subtests 

Test Summary Report
-------------------
exercises/grains/.meta/solutions/grains.t (Wstat: 512 Tests: 12 Failed: 2)
  Failed tests:  8, 12
  Non-zero exit status: 2
Files=1, Tests=12,  0 wallclock secs ( 0.02 usr  0.00 sys +  0.10 cusr  0.01 csys =  0.13 CPU)
Result: FAIL

Also, these changes result in doing a string comparison rather than a numeric one. Do you have an example/link to a solution where issues have come up for this?

[kappa]

The problem to solve is #239. There are examples of problematic solutions that pass the tests there. That issue is old and it had been fixed at one point but then regressed and is relevant for the current master.

My change removes BigFloat from the test code and relies on the Grains.pm having "use bigint" and always returning bigints -- and now I see the problem, this is not present in the stub which I incorrectly assumed.

[m-dango]

Ah yes I see what's happened.

use bigint was intentionally excluded from the stub. The exercise is quite straightforward without it, and I sense that leaving it in would have people miss that it is there and what it is for.

[m-dango]

Thanks @kappa!